--- title: Learn About AI Ontology | Ontollo description: Learn why an ontology is critical to ensure AI is accurate and cost-effective. Compare it with other approaches, including context dumps and RAG to see the difference. --- [← Back to Ontollo](https://ontollo.com/?hsLang=en) ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) The Role of Ontology In AI # AI is making your operation faster. Is it making it right? A 5-minute walkthrough on why point AI tools speed up the wrong decisions, and how grounding AI in one connected picture of your business fixes accuracy and AI spend at the same time. ~5 min · tailored to your industry · sources cited **Start here.** Pick your industry and we'll build every example around it. Construction Jobs, crews, equipment, and costs across every site Energy Substations, spares, and switching crews Health Care Operations Care workers, clients, and visits Agriculture Illustrative industry example; content pending client review Choose one to begin ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) Your industry Construction Energy Health Care Operations Agriculture Industry**Change** Request a demo ⌄ Talk with Ontollo ## Why AI gets operations wrong, and how to fix it. 1**The speed trap**Every tool sees a slice Point AI sees a slice Copilot checks its workspace What Copilot could not see The cost of a confident guess 2**Context matters**Three ways to give AI context Three approaches Context Dump RAG OAG How they stack up 3**What it costs at scale**People, questions and tokens 4**Where Ontollo fits**One connected picture Disconnected systems System of Reality Decide and act Portfolio in context 5**Book a demo**See your data connected 6**FAQ**Questions and definitions Frequently asked questions Four definitions 1 · The speed trap # Every tool you run now has AI. Each one sees a slice. Your fleet system, your scheduler, your HR platform: they all ship an AI assistant now. Each one is smart about its own data and blind to everyone else's. So when it acts, it acts fast, on a partial picture. It's the old story of the blind men and the elephant. One grabs the trunk, one grabs the leg, and each describes a different animal. Point AI tools do the same thing. They just do it at machine speed. Project management AI Sees Cedar Ridge Tower needs a crane Thursday Blind spotThat you already own one Equipment telematics AI Sees An owned crane idle two sites over Blind spotWhich job needs it Thursday HR & certifications AI Sees Who's certified to run that crane Blind spotWhich of them is free Thursday Procurement & finance AI Sees A $6,600 external crane hire Blind spotThat an owned crane is sitting idle 1 · The speed trap # Ask Copilot. Get an answer fast. An illustrative example of a question answered from one connected workspace, not the complete operational picture. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg) **Microsoft 365 Copilot** Is Cedar Ridge Tower ready for Thursday's crane lift? Looking through connected Microsoft 365 content… **Everything looks on track.** The schedule, the crane hire quote, and the crew notes show no blockers. Based only on connected Microsoft 365 content │ Ask anything↑ 1 · The speed trap # One answer. One missing delivery. The AI was quick, but the shipping information was outside the systems it checked. **FedEx** External logistics **Shipment delayed** The rigging for Thursday's crane lift won't arrive in time. Not available to the assistant's workspace search The AI made the call faster. It also made it wrong faster. 2 · Context is the whole game # Context Dump You attach files to an AI assistant such as Claude. It reads the information and makes a best guess based on what was provided. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Cedar Ridge Tower needs a crane Thursday. Can a crane we own replace the $6,600 hire, and do we have a certified operator free to run it? Copilot ### Redeploy the idle crane. Its last operator can run it. The equipment list shows an owned crane idle two sites over, and the operator who ran it last month is certified. Move both and skip the $6,600 hire. Tokens per question**~8,000** Message Copilot ↑ Incorrect ### Picks an operator whose card expired Every equipment list, roster, and certificate goes in raw. It pairs the crane with last month's operator without catching that his crane card expired last week, or that he's booked on another job Thursday. 2 · Context is the whole game # Retrieval-Augmented Generation (RAG) You connect an AI assistant to tools and datasets you use. It retrieves relevant information and attempts to match it to your question. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Cedar Ridge Tower needs a crane Thursday. Can a crane we own replace the $6,600 hire, and do we have a certified operator free to run it? Copilot ### An owned crane is idle two sites over. Telematics shows it idle this week. I couldn't confirm a certified operator who's free Thursday, so you may still need the $6,600 hire. Tokens per question**~1,300** Message Copilot ↑ !Partial ### Finds the crane, misses the operator A telematics lookup finds the idle crane cleanly. Credit where it's due: that's a real query. But a certified operator who's free Thursday lives in HR and the schedule, and two separate lookups don't tell the AI they're about the same job. 2 · Context is the whole game # Ontology-Augmented Generation (OAG) OAG uses your business ontology—a framework of domain concepts, explicit relationships, business logic, constraints, and operational data schemas—as context. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg)Ontology-grounded assistant You Cedar Ridge Tower needs a crane Thursday. Can a crane we own replace the $6,600 hire, and do we have a certified operator free to run it? Ontollo ### Skip the $6,600 hire. Redeploy the crane you own. The owned crane two sites over is idle per telematics. One operator holds a current crane card and has no assignment Thursday. I drafted the move for your approval. Procore stays the record for the commitment. Tokens per question**~500** Message Ontollo ↑ ✓Correct,sourced ### Names the crane and the operator, sourced One query walks job, crane, telematics, operator, and certification together. You get the idle crane and a certified operator who's actually free, with a source for every fact. Source · System of recordSource · Field update 3 · What it costs at scale # The right answer is also the cheap one. Every word the AI reads costs you tokens. Compare how the same question scales across your team. Employees using AI 35 people 1 employee300 employees **35 prompts per employee per day** 1,225 total prompts/day. Estimated costs use an illustrative input-token price. Microsoft 365 Copilot seat subscriptions are billed differently. Tokens per month Context Dump 294.0M RAG 47.8M OAG 18.4M Context Dump**$882/mo** RAG**$143/mo** OAG**$55/mo** Illustrative estimates assume 30 days per month and count input tokens only. Output tokens, caching, licensing, and other fees are excluded. 4 · Where Ontollo fits # Your tools already know a lot. They don't know each other. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo-white.svg)One live operational picture ![Procore](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-procore.svg) ![Autodesk](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-autodesk.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) **Coupa** ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) Ontollo adds the context layer across all your tools, providing AI with a live picture of your business to reason over and act on. 4 · Where Ontollo fits # Give every AI tool the same picture of your business. Think control tower, not another dashboard. Events come in from the field, the full picture is already on the screen, a person makes the call, and the action goes back out. At Ontollo, we build that picture once, and every app, agent, and person works from it. Your systems of record ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) Your project management, ERP, procurement, equipment, and HR systems. You keep them. Ontollo connects them. System of Reality ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) One current, connected picture of your jobs, crews, equipment, certifications, and costs, built on an ontology: a shared rulebook for what everything is and how it connects. System of Action ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) AI prepares the next step with every relevant fact attached. A person confirms. The change lands back in the system you already use, and every decision gets recorded. ✓ **Keep your tools**Ontollo sits on top of what you run today. No rip and replace. ✓ **Value in weeks**No six-month data platform project before anyone sees a result. We model only what your operations workflows need. ✓ **People stay in charge**AI reads, matches, and flags. Your team approves before anything changes. ✓ **One set of rules for every builder**Same data, same governance, whether it's an app, an agent, or someone on your team building their own workflow. 2 · Context is the whole game # There are three ways to give AI context about your business. Only one connects the dots. An AI answer is only as good as what the AI can see when it answers. Here's how the three common approaches stack up, in plain terms. Approach 1 ### Context Dump You paste everything into the prompt and hope. The AI skims, samples, and guesses. You pay for every page it reads, and the bill grows every time your data does. Approach 2 ### RAG The AI searches your documents or queries one system, then answers from what it found. Cheaper, and great for a single lookup. But it can't tell that two files describe the same job, crew, or asset. Approach 3 ### OAG (ontology-grounded) Every relevant fact is already connected, current, and sourced before the AI says a word. It asks one precise question of a connected map of your business, and its job shrinks to explaining the answer. 1 · The speed trap # A confident answer isn't the same as a correct answer. 22–94%of AI guesses are wrong When AI models lack an answer, they guess. The worst part? They don't tell you they're guessing. Stanford AI Index 2026, ch. 3; NIST AI 600-1 2 · Context is the whole game # How the three approaches stack up. ← Swipe to compare all approaches → | | Context Dump | RAG | OAG | | --- | --- | --- | --- | | Cross-system questions | ×Often wrong, confidently | —Right on one system, misses the join | ✓Built for them | | Cost per question | ×Highest | —Moderate | ✓Lowest | | As your data grows | ×Cost grows with it | —Cost stays capped, accuracy slips | ✓Cost stays flat | | Can you trace the answer? | ×No | —Partly | ✓Every fact has a source | | Safe to act on? | ×Not without redoing the work | —With heavy checking | ✓A person confirms, then it acts | 4 · Where Ontollo fits # Decide and act with confidence. Stream · Job 1931 True exposure and why the forecast moved Exposure Engine · Riverside PlazaMargin risk Ask below, then send — Ontollo drafts the moves; you keep the call. TN Tiffany McIntyreProject Manager · Job 19316:58 AM Exposure Engineworking Getting the answer Exposure Engineclose the loop6:59 AM `/task, draft the moves`ran Two actions clear most of the gap. The system does the typing; you keep the call. 1 Post #INV-61704 · $412,600 to #RiversidePlaza · CC 04-160 · #PO-3847 Three-way match: PO, L5 receiving log, invoice. Coding inherited from the PO. pilot 98% 2 Resolve duplicate: INV-61688 vs manual accrual JE-1840-031 Same vendor, amount and scope. Keep the invoice, reverse the accrual; count once. 94% Confirm & run Adjust Confirmed, running Reply in your own words… **@person** **#object** **/skill** 4 · Where Ontollo fits # A live view across your portfolio. Stream Company Projects Compliance Map TN New project Filter projects… Views Portfolio On track**2** At risk**3** Jobs *· 5* Riverside Plaza**71** Bayport Data Hall 2**76** North Bay Expansion**74** Med Center Pavilion**91** Katy Logistics DC**93** Search jobs, POs, people… TN Projects · Portfolio Every job’s path to its completion cost What’s posted, what’s committed but not yet visible to Dynamics, and where the forecast lands against budget. **2**On track · ≥80% confidence **3**At risk · 65 to 79% **0**Slipping · below 65% **$34.8M**Invisible to Dynamics Path to completion cost · *BUDGET tick · POSTED solid · COMMITTED ◇ · FORECAST EAC ●* Riverside Plaza Portland · Structural · $41.8M BUDGET EAC on budget, margin thinning: +$385K this week **71%**CONF Bayport Data Hall 2 Seattle · Structural · $33.4M BUDGET+$0.7M +$0.7M, retainage tail widening **76%**CONF North Bay Expansion Portland · Structural · $16.4M BUDGET+$0.5M +$0.5M, rebar rate drifting **74%**CONF Med Center Pavilion Seattle · Structural · $28.3M BUDGET On plan, under budget **91%**CONF 5 · Book a demo # See this on your own construction data. Bring us one question like this one that your team answers by hand today. We'll show you what it looks like when the answer comes back connected, sourced, and ready for a person to act on. 6 · What your team will ask # The concerns are fair. Here's how each one gets handled. ### Is our data safe? Public AI tools can keep what you type, and some train on it. Company data stays in approved, access-controlled systems, with each person seeing only their slice. ### Will it give us wrong answers? On its own, AI predicts from patterns, so it can be confidently wrong. Every answer traces back to a source, and a person checks it before it counts. ### Our data is a mess. AI is only as good as the records it reads. Capture data once, where the work happens, and link it across systems instead of cleaning it forever. ### Will AI make decisions for us? It can only do what you give it permission to do. You set the boundaries. AI prepares the call. People make it. 6 · What your team will ask # Want the definitions? Four terms, in plain language ##### Ontology A shared rulebook for your business: what counts as an asset, a site, a crew, or a work order, and how they connect. Every system you run keeps its own private version. An ontology is the one everyone agrees on. [Tom Gruber, Knowledge Acquisition, 1993](https://tomgruber.org/writing/definition-of-ontology) ##### Knowledge graph Your real data mapped onto that rulebook: this machine, at this site, since this date. A web of connected facts a computer can walk, instead of searching blindly. [Hogan et al., ACM Computing Surveys, 2021](https://doi.org/10.1145/3447772) ##### RAG (retrieval-augmented generation) The AI searches documents or queries a system first, then answers from the handful of results that look most relevant. Good for one lookup in one place. [Lewis et al., NeurIPS, 2020](https://arxiv.org/abs/2005.11401) ##### OAG (ontology-augmented generation) The AI queries the knowledge graph directly and gets back connected, sourced facts. The facts come from the graph, not from the model's best guess, and the model's job is to explain them. [Related research on graph-grounded retrieval: Edge et al. (Microsoft Research), 2024](https://arxiv.org/abs/2404.16130) 1 · The speed trap # Every tool you run now has AI. Each one sees a slice. Your fleet system, your scheduler, your HR platform: they all ship an AI assistant now. Each one is smart about its own data and blind to everyone else's. So when it acts, it acts fast, on a partial picture. It's the old story of the blind men and the elephant. One grabs the trunk, one grabs the leg, and each describes a different animal. Point AI tools do the same thing. They just do it at machine speed. Work management AI Sees The Eastfield upgrade needs a crew Tuesday Blind spotWhether that crew is authorized for the job Asset / GIS AI Sees Spare transformers and their ratings Blind spotWhich crew is going to install one Training records AI Sees Who holds a switching authorization Blind spotWhich voltage class the job actually is Compliance AI Sees Which authorizations have lapsed Blind spotThat one of them is on Tuesday's roster 1 · The speed trap # Ask Copilot. Get an answer fast. An illustrative example of a question answered from one connected workspace, not the complete operational picture. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg) **Microsoft 365 Copilot** Is the Eastfield switching job on track for Tuesday? Looking through connected Microsoft 365 content… **The job looks ready.** The work order and crew notes in the files show no blockers. Based only on connected Microsoft 365 content │ Ask anything↑ 1 · The speed trap # One answer. One equipment conflict. The AI saw the work order, but not the equipment committed elsewhere. **Equipment allocation** Separate asset system **Equipment unavailable** The required equipment is still assigned to another project. Not available to the assistant's workspace search The work looked ready. The equipment was not. 2 · Context is the whole game # Context Dump You attach files to an AI assistant such as Claude. It reads the information and makes a best guess based on what was provided. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Which of our spare transformers can go to the Eastfield substation upgrade, and is the crew's switching authorization still current for that voltage class? Copilot Tokens per question**~9,500** Message Copilot ↑ Incorrect ### Lists a lapsed authorization as current Every transformer spec and training record goes in raw. An authorization that lapsed last month still shows as valid, because nothing forces a check against today's date. 2 · Context is the whole game # Retrieval-Augmented Generation (RAG) You connect an AI assistant to tools and datasets you use. It retrieves relevant information and attempts to match it to your question. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Which of our spare transformers can go to the Eastfield substation upgrade, and is the crew's switching authorization still current for that voltage class? Copilot Tokens per question**~1,350** Message Copilot ↑ !Partial ### Finds the transformer, misses the match A lookup in the asset system finds the transformer and its rating. A document search surfaces the authorization too. Neither checks one against the other, and whether the authorization covers this voltage class is a cross-system fact no single query returns. 2 · Context is the whole game # Ontology-Augmented Generation (OAG) OAG uses your business ontology—a framework of domain concepts, explicit relationships, business logic, constraints, and operational data schemas—as context. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg)Ontology-grounded assistant You Which of our spare transformers can go to the Eastfield substation upgrade, and is the crew's switching authorization still current for that voltage class? Ontollo Tokens per question**~550** Message Ontollo ↑ ✓Correct,sourced ### Rating, authorization, and expiry confirmed together One query confirms the rating matches, the authorization is current as of today, and the voltage class lines up. Every fact sourced, because a wrong answer here is a safety incident. Source · System of recordSource · Field update 3 · What it costs at scale # The right answer is also the cheap one. Every word the AI reads costs you tokens. Compare how the same question scales across your team. Employees using AI 35 people 1 employee300 employees **35 prompts per employee per day** 1,225 total prompts/day. Estimated costs use an illustrative input-token price. Microsoft 365 Copilot seat subscriptions are billed differently. Tokens per month Context Dump 349.1M RAG 49.6M OAG 20.2M Context Dump**$1,047/mo** RAG**$149/mo** OAG**$61/mo** Illustrative estimates assume 30 days per month and count input tokens only. Output tokens, caching, licensing, and other fees are excluded. 4 · Where Ontollo fits # Your tools already know a lot. They don't know each other. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo-white.svg)One live operational picture ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) ![IBM](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-ibm.svg) **SCADA** **GIS** Ontollo adds the context layer across all your tools, providing AI with a live picture of your business to reason over and act on. 4 · Where Ontollo fits # Give every AI tool the same picture of your business. Think control tower, not another dashboard. Events come in from the field, the full picture is already on the screen, a person makes the call, and the action goes back out. At Ontollo, we build that picture once, and every app, agent, and person works from it. Your systems of record ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) Your asset/GIS, work management, training, and compliance systems. You keep them. Ontollo connects them. System of Reality ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) One current, connected picture of your assets, crews, authorizations, and work orders, built on an ontology: a shared rulebook for what everything is and how it connects. System of Action ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) AI prepares the next step with every relevant fact attached. A person confirms. The change lands back in the system you already use, and every decision gets recorded. ✓ **Keep your tools**Ontollo sits on top of what you run today. No rip and replace. ✓ **Value in weeks**No six-month data platform project before anyone sees a result. We model only what your operations workflows need. ✓ **People stay in charge**AI reads, matches, and flags. Your team approves before anything changes. ✓ **One set of rules for every builder**Same data, same governance, whether it's an app, an agent, or someone on your team building their own workflow. 2 · Context is the whole game # There are three ways to give AI context about your business. Only one connects the dots. An AI answer is only as good as what the AI can see when it answers. Here's how the three common approaches stack up, in plain terms. Approach 1 ### Context Dump You paste everything into the prompt and hope. The AI skims, samples, and guesses. You pay for every page it reads, and the bill grows every time your data does. Approach 2 ### RAG The AI searches your documents or queries one system, then answers from what it found. Cheaper, and great for a single lookup. But it can't tell that two files describe the same job, crew, or asset. Approach 3 ### OAG (ontology-grounded) Every relevant fact is already connected, current, and sourced before the AI says a word. It asks one precise question of a connected map of your business, and its job shrinks to explaining the answer. 1 · The speed trap # A confident answer isn't the same as a correct answer. 22–94%of AI guesses are wrong When AI models lack an answer, they guess. The worst part? They don't tell you they're guessing. Stanford AI Index 2026, ch. 3; NIST AI 600-1 2 · Context is the whole game # How the three approaches stack up. ← Swipe to compare all approaches → | | Context Dump | RAG | OAG | | --- | --- | --- | --- | | Cross-system questions | ×Often wrong, confidently | —Right on one system, misses the join | ✓Built for them | | Cost per question | ×Highest | —Moderate | ✓Lowest | | As your data grows | ×Cost grows with it | —Cost stays capped, accuracy slips | ✓Cost stays flat | | Can you trace the answer? | ×No | —Partly | ✓Every fact has a source | | Safe to act on? | ×Not without redoing the work | —With heavy checking | ✓A person confirms, then it acts | 4 · Where Ontollo fits # Decide and act with confidence. Stream · Job 1931 True exposure and why the forecast moved Exposure Engine · Riverside PlazaMargin risk Ask below, then send — Ontollo drafts the moves; you keep the call. TN Tiffany McIntyreProject Manager · Job 19316:58 AM Exposure Engineworking Getting the answer Exposure Engineclose the loop6:59 AM `/task, draft the moves`ran Two actions clear most of the gap. The system does the typing; you keep the call. 1 Post #INV-61704 · $412,600 to #RiversidePlaza · CC 04-160 · #PO-3847 Three-way match: PO, L5 receiving log, invoice. Coding inherited from the PO. pilot 98% 2 Resolve duplicate: INV-61688 vs manual accrual JE-1840-031 Same vendor, amount and scope. Keep the invoice, reverse the accrual; count once. 94% Confirm & run Adjust Confirmed, running Reply in your own words… **@person** **#object** **/skill** 4 · Where Ontollo fits # A live view across your portfolio. Stream Company Projects Compliance Map TN New project Filter projects… Views Portfolio On track**2** At risk**3** Jobs *· 5* Riverside Plaza**71** Bayport Data Hall 2**76** North Bay Expansion**74** Med Center Pavilion**91** Katy Logistics DC**93** Search jobs, POs, people… TN Projects · Portfolio Every job’s path to its completion cost What’s posted, what’s committed but not yet visible to Dynamics, and where the forecast lands against budget. **2**On track · ≥80% confidence **3**At risk · 65 to 79% **0**Slipping · below 65% **$34.8M**Invisible to Dynamics Path to completion cost · *BUDGET tick · POSTED solid · COMMITTED ◇ · FORECAST EAC ●* Riverside Plaza Portland · Structural · $41.8M BUDGET EAC on budget, margin thinning: +$385K this week **71%**CONF Bayport Data Hall 2 Seattle · Structural · $33.4M BUDGET+$0.7M +$0.7M, retainage tail widening **76%**CONF North Bay Expansion Portland · Structural · $16.4M BUDGET+$0.5M +$0.5M, rebar rate drifting **74%**CONF Med Center Pavilion Seattle · Structural · $28.3M BUDGET On plan, under budget **91%**CONF 5 · Book a demo # See this on your own Energy data. Bring us one question like this one that your team answers by hand today. We'll show you what it looks like when the answer comes back connected, sourced, and ready for a person to act on. 6 · What your team will ask # The concerns are fair. Here's how each one gets handled. ### Is our data safe? Public AI tools can keep what you type, and some train on it. Company data stays in approved, access-controlled systems, with each person seeing only their slice. ### Will it give us wrong answers? On its own, AI predicts from patterns, so it can be confidently wrong. Every answer traces back to a source, and a person checks it before it counts. ### Our data is a mess. AI is only as good as the records it reads. Capture data once, where the work happens, and link it across systems instead of cleaning it forever. ### Will AI make decisions for us? It can only do what you give it permission to do. You set the boundaries. AI prepares the call. People make it. 6 · What your team will ask # Want the definitions? Four terms, in plain language ##### Ontology A shared rulebook for your business: what counts as an asset, a site, a crew, or a work order, and how they connect. Every system you run keeps its own private version. An ontology is the one everyone agrees on. [Tom Gruber, Knowledge Acquisition, 1993](https://tomgruber.org/writing/definition-of-ontology) ##### Knowledge graph Your real data mapped onto that rulebook: this machine, at this site, since this date. A web of connected facts a computer can walk, instead of searching blindly. [Hogan et al., ACM Computing Surveys, 2021](https://doi.org/10.1145/3447772) ##### RAG (retrieval-augmented generation) The AI searches documents or queries a system first, then answers from the handful of results that look most relevant. Good for one lookup in one place. [Lewis et al., NeurIPS, 2020](https://arxiv.org/abs/2005.11401) ##### OAG (ontology-augmented generation) The AI queries the knowledge graph directly and gets back connected, sourced facts. The facts come from the graph, not from the model's best guess, and the model's job is to explain them. [Related research on graph-grounded retrieval: Edge et al. (Microsoft Research), 2024](https://arxiv.org/abs/2404.16130) 1 · The speed trap # Every tool you run now has AI. Each one sees a slice. Your fleet system, your scheduler, your HR platform: they all ship an AI assistant now. Each one is smart about its own data and blind to everyone else's. So when it acts, it acts fast, on a partial picture. It's the old story of the blind men and the elephant. One grabs the trunk, one grabs the leg, and each describes a different animal. Point AI tools do the same thing. They just do it at machine speed. Rostering AI Sees Which carers are free and nearby Blind spotWhat this client's care plan requires Client management AI Sees Today's visit, window, and address Blind spotWho's qualified to cover it Credentialing AI Sees Police clearances and certifications Blind spotWhich ones expired this month Care plans AI Sees This client needs a two-person lift Blind spotWho's been rostered to the visit 1 · The speed trap # Ask Copilot. Get an answer fast. An illustrative example of a question answered from one connected workspace, not the complete operational picture. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg) **Microsoft 365 Copilot** Is today's home visit still covered? Looking through connected Microsoft 365 content… **The visit looks covered.** The roster file shows a worker assigned. No cancellation appears in the team messages. Based only on connected Microsoft 365 content │ Ask anything↑ 1 · The speed trap # One answer. One unavailable care team. The AI saw the scheduled visit, but missed a staffing change. **Care staffing roster** Separate care operations system **Care team unavailable** The assigned care team member cannot cover the planned visit. Not available to the assistant's workspace search The visit looked covered. The care team was unavailable. 2 · Context is the whole game # Context Dump You attach files to an AI assistant such as Claude. It reads the information and makes a best guess based on what was provided. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Which available care worker can cover today's home visit, and do they hold a current police clearance and the manual handling certification this client's care plan requires? Copilot Tokens per question**~7,000** Message Copilot ↑ Incorrect ### Rosters a worker whose clearance lapsed Every availability record, credential, and care plan goes in raw. It picks a worker who looks free and close, without catching that their police clearance expired last month. 2 · Context is the whole game # Retrieval-Augmented Generation (RAG) You connect an AI assistant to tools and datasets you use. It retrieves relevant information and attempts to match it to your question. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Which available care worker can cover today's home visit, and do they hold a current police clearance and the manual handling certification this client's care plan requires? Copilot Tokens per question**~1,200** Message Copilot ↑ !Partial ### Finds a worker, misses the care plan A lookup in rostering finds an available, nearby worker. A search surfaces their credentials. Neither checks this client's specific requirement, like a two-person lift certification, against what the worker actually holds. 2 · Context is the whole game # Ontology-Augmented Generation (OAG) OAG uses your business ontology—a framework of domain concepts, explicit relationships, business logic, constraints, and operational data schemas—as context. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg)Ontology-grounded assistant You Which available care worker can cover today's home visit, and do they hold a current police clearance and the manual handling certification this client's care plan requires? Ontollo Tokens per question**~475** Message Ontollo ↑ ✓Correct,sourced ### Worker, clearance, and care plan confirmed together One query confirms the worker is free, their clearance and certification are current, and both match this client's care plan. Every fact sourced, because a miss here is a safeguarding issue. Source · System of recordSource · Field update 3 · What it costs at scale # The right answer is also the cheap one. Every word the AI reads costs you tokens. Compare how the same question scales across your team. Employees using AI 35 people 1 employee300 employees **35 prompts per employee per day** 1,225 total prompts/day. Estimated costs use an illustrative input-token price. Microsoft 365 Copilot seat subscriptions are billed differently. Tokens per month Context Dump 257.3M RAG 44.1M OAG 17.5M Context Dump**$772/mo** RAG**$132/mo** OAG**$52/mo** Illustrative estimates assume 30 days per month and count input tokens only. Output tokens, caching, licensing, and other fees are excluded. 4 · Where Ontollo fits # Your tools already know a lot. They don't know each other. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo-white.svg)One live operational picture **Epic** ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) **EHR** **Scheduling** Ontollo adds the context layer across all your tools, providing AI with a live picture of your business to reason over and act on. 4 · Where Ontollo fits # Give every AI tool the same picture of your business. Think control tower, not another dashboard. Events come in from the field, the full picture is already on the screen, a person makes the call, and the action goes back out. At Ontollo, we build that picture once, and every app, agent, and person works from it. Your systems of record ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) Your rostering, client management, credentialing, and care plan systems. You keep them. Ontollo connects them. System of Reality ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) One current, connected picture of your workers, clients, visits, and credentials, built on an ontology: a shared rulebook for what everything is and how it connects. System of Action ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) AI prepares the next step with every relevant fact attached. A person confirms. The change lands back in the system you already use, and every decision gets recorded. ✓ **Keep your tools**Ontollo sits on top of what you run today. No rip and replace. ✓ **Value in weeks**No six-month data platform project before anyone sees a result. We model only what your operations workflows need. ✓ **People stay in charge**AI reads, matches, and flags. Your team approves before anything changes. ✓ **One set of rules for every builder**Same data, same governance, whether it's an app, an agent, or someone on your team building their own workflow. 2 · Context is the whole game # There are three ways to give AI context about your business. Only one connects the dots. An AI answer is only as good as what the AI can see when it answers. Here's how the three common approaches stack up, in plain terms. Approach 1 ### Context Dump You paste everything into the prompt and hope. The AI skims, samples, and guesses. You pay for every page it reads, and the bill grows every time your data does. Approach 2 ### RAG The AI searches your documents or queries one system, then answers from what it found. Cheaper, and great for a single lookup. But it can't tell that two files describe the same job, crew, or asset. Approach 3 ### OAG (ontology-grounded) Every relevant fact is already connected, current, and sourced before the AI says a word. It asks one precise question of a connected map of your business, and its job shrinks to explaining the answer. 1 · The speed trap # A confident answer isn't the same as a correct answer. 22–94%of AI guesses are wrong When AI models lack an answer, they guess. The worst part? They don't tell you they're guessing. Stanford AI Index 2026, ch. 3; NIST AI 600-1 2 · Context is the whole game # How the three approaches stack up. ← Swipe to compare all approaches → | | Context Dump | RAG | OAG | | --- | --- | --- | --- | | Cross-system questions | ×Often wrong, confidently | —Right on one system, misses the join | ✓Built for them | | Cost per question | ×Highest | —Moderate | ✓Lowest | | As your data grows | ×Cost grows with it | —Cost stays capped, accuracy slips | ✓Cost stays flat | | Can you trace the answer? | ×No | —Partly | ✓Every fact has a source | | Safe to act on? | ×Not without redoing the work | —With heavy checking | ✓A person confirms, then it acts | 4 · Where Ontollo fits # Decide and act with confidence. Stream · Job 1931 True exposure and why the forecast moved Exposure Engine · Riverside PlazaMargin risk Ask below, then send — Ontollo drafts the moves; you keep the call. TN Tiffany McIntyreProject Manager · Job 19316:58 AM Exposure Engineworking Getting the answer Exposure Engineclose the loop6:59 AM `/task, draft the moves`ran Two actions clear most of the gap. The system does the typing; you keep the call. 1 Post #INV-61704 · $412,600 to #RiversidePlaza · CC 04-160 · #PO-3847 Three-way match: PO, L5 receiving log, invoice. Coding inherited from the PO. pilot 98% 2 Resolve duplicate: INV-61688 vs manual accrual JE-1840-031 Same vendor, amount and scope. Keep the invoice, reverse the accrual; count once. 94% Confirm & run Adjust Confirmed, running Reply in your own words… **@person** **#object** **/skill** 4 · Where Ontollo fits # A live view across your portfolio. Stream Company Projects Compliance Map TN New project Filter projects… Views Portfolio On track**2** At risk**3** Jobs *· 5* Riverside Plaza**71** Bayport Data Hall 2**76** North Bay Expansion**74** Med Center Pavilion**91** Katy Logistics DC**93** Search jobs, POs, people… TN Projects · Portfolio Every job’s path to its completion cost What’s posted, what’s committed but not yet visible to Dynamics, and where the forecast lands against budget. **2**On track · ≥80% confidence **3**At risk · 65 to 79% **0**Slipping · below 65% **$34.8M**Invisible to Dynamics Path to completion cost · *BUDGET tick · POSTED solid · COMMITTED ◇ · FORECAST EAC ●* Riverside Plaza Portland · Structural · $41.8M BUDGET EAC on budget, margin thinning: +$385K this week **71%**CONF Bayport Data Hall 2 Seattle · Structural · $33.4M BUDGET+$0.7M +$0.7M, retainage tail widening **76%**CONF North Bay Expansion Portland · Structural · $16.4M BUDGET+$0.5M +$0.5M, rebar rate drifting **74%**CONF Med Center Pavilion Seattle · Structural · $28.3M BUDGET On plan, under budget **91%**CONF 5 · Book a demo # See this on your own health care data. Bring us one question like this one that your team answers by hand today. We'll show you what it looks like when the answer comes back connected, sourced, and ready for a person to act on. 6 · What your team will ask # The concerns are fair. Here's how each one gets handled. ### Is our data safe? Public AI tools can keep what you type, and some train on it. Company data stays in approved, access-controlled systems, with each person seeing only their slice. ### Will it give us wrong answers? On its own, AI predicts from patterns, so it can be confidently wrong. Every answer traces back to a source, and a person checks it before it counts. ### Our data is a mess. AI is only as good as the records it reads. Capture data once, where the work happens, and link it across systems instead of cleaning it forever. ### Will AI make decisions for us? It can only do what you give it permission to do. You set the boundaries. AI prepares the call. People make it. 6 · What your team will ask # Want the definitions? Four terms, in plain language ##### Ontology A shared rulebook for your business: what counts as an asset, a site, a crew, or a work order, and how they connect. Every system you run keeps its own private version. An ontology is the one everyone agrees on. [Tom Gruber, Knowledge Acquisition, 1993](https://tomgruber.org/writing/definition-of-ontology) ##### Knowledge graph Your real data mapped onto that rulebook: this machine, at this site, since this date. A web of connected facts a computer can walk, instead of searching blindly. [Hogan et al., ACM Computing Surveys, 2021](https://doi.org/10.1145/3447772) ##### RAG (retrieval-augmented generation) The AI searches documents or queries a system first, then answers from the handful of results that look most relevant. Good for one lookup in one place. [Lewis et al., NeurIPS, 2020](https://arxiv.org/abs/2005.11401) ##### OAG (ontology-augmented generation) The AI queries the knowledge graph directly and gets back connected, sourced facts. The facts come from the graph, not from the model's best guess, and the model's job is to explain them. [Related research on graph-grounded retrieval: Edge et al. (Microsoft Research), 2024](https://arxiv.org/abs/2404.16130) 1 · The speed trap # Every tool you run now has AI. Each one sees a slice. Your fleet system, your scheduler, your HR platform: they all ship an AI assistant now. Each one is smart about its own data and blind to everyone else's. So when it acts, it acts fast, on a partial picture. It's the old story of the blind men and the elephant. One grabs the trunk, one grabs the leg, and each describes a different animal. Point AI tools do the same thing. They just do it at machine speed. Fleet system AI Sees Three excavators sitting idle Blind spotWhich one is overdue for service Project system AI Sees Stage 2 needs an excavator Monday Blind spotWhich machines are actually free HR & certifications AI Sees Who's ticketed to operate one Blind spotWhere those operators are this week Finance AI Sees What the external hire costs Blind spotThat owned machines are sitting idle 1 · The speed trap # Ask Copilot. Get an answer fast. An illustrative example of a question answered from one connected workspace, not the complete operational picture. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg) **Microsoft 365 Copilot** Is today's harvest block still covered? Looking through connected Microsoft 365 content… **The crew looks assigned.** The schedule in the shared files shows a crew on that block. No change appears in the team messages. Based only on connected Microsoft 365 content │ Ask anything↑ 1 · The speed trap # One answer. One unavailable crew. The AI saw the schedule, but not the latest field crew availability. **Field crew roster** Separate operations system **Crew unavailable** The assigned crew is unavailable for the planned work. Not available to the assistant's workspace search The work was scheduled. The crew was not available. 2 · Context is the whole game # Context Dump You attach files to an AI assistant such as Claude. It reads the information and makes a best guess based on what was provided. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Which field crew can cover today's harvest block, and is their equipment certification still current for the machine on that block? Copilot Tokens per question**~7,500** Message Copilot ↑ Incorrect ### Sends a crew whose ticket lapsed Every roster line and certification goes in raw. It picks a crew that looks free, without catching that their equipment ticket expired. 2 · Context is the whole game # Retrieval-Augmented Generation (RAG) You connect an AI assistant to tools and datasets you use. It retrieves relevant information and attempts to match it to your question. ![Microsoft Copilot](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/learn/microsoft-copilot.svg)Microsoft 365 Copilot You Which field crew can cover today's harvest block, and is their equipment certification still current for the machine on that block? Copilot Tokens per question**~1,250** Message Copilot ↑ !Partial ### Finds a crew, misses the machine ticket A lookup in the roster finds an available crew. A search surfaces their certification. Neither checks that ticket against the machine on today's block. 2 · Context is the whole game # Ontology-Augmented Generation (OAG) OAG uses your business ontology—a framework of domain concepts, explicit relationships, business logic, constraints, and operational data schemas—as context. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg)Ontology-grounded assistant You Which field crew can cover today's harvest block, and is their equipment certification still current for the machine on that block? Ontollo Tokens per question**~480** Message Ontollo ↑ ✓Correct,sourced ### Crew, certification, and machine confirmed together One query confirms the crew is free, their certification is current, and it covers the machine on that block. Every fact sourced. Source · System of recordSource · Field update 3 · What it costs at scale # The right answer is also the cheap one. Every word the AI reads costs you tokens. Compare how the same question scales across your team. Employees using AI 35 people 1 employee300 employees **35 prompts per employee per day** 1,225 total prompts/day. Estimated costs use an illustrative input-token price. Microsoft 365 Copilot seat subscriptions are billed differently. Tokens per month Context Dump 275.6M RAG 45.9M OAG 17.6M Context Dump**$827/mo** RAG**$138/mo** OAG**$53/mo** Illustrative estimates assume 30 days per month and count input tokens only. Output tokens, caching, licensing, and other fees are excluded. 4 · Where Ontollo fits # Your tools already know a lot. They don't know each other. ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo-white.svg)One live operational picture ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) **John Deere** **GIS** **ERP** Ontollo adds the context layer across all your tools, providing AI with a live picture of your business to reason over and act on. 4 · Where Ontollo fits # Give every AI tool the same picture of your business. Think control tower, not another dashboard. Events come in from the field, the full picture is already on the screen, a person makes the call, and the action goes back out. At Ontollo, we build that picture once, and every app, agent, and person works from it. Your systems of record ![SAP](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-sap.svg) ![Oracle](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-oracle.svg) ![Dynamics 365](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-logos/logo-microsoft-dynamics-365.svg) Your fleet, project, HR, and finance systems. You keep them. Ontollo connects them. System of Reality ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) One current, connected picture of your people, machines, projects, and certifications, built on an ontology: a shared rulebook for what everything is and how it connects. System of Action ![Ontollo](https://www.ontollo.com/hubfs/raw_assets/public/Ontollo%20x%20Clean/images/ontollo-pov/Ontollo_logo_fullcolor-100px.svg) AI prepares the next step with every relevant fact attached. A person confirms. The change lands back in the system you already use, and every decision gets recorded. ✓ **Keep your tools**Ontollo sits on top of what you run today. No rip and replace. ✓ **Value in weeks**No six-month data platform project before anyone sees a result. We model only what your operations workflows need. ✓ **People stay in charge**AI reads, matches, and flags. Your team approves before anything changes. ✓ **One set of rules for every builder**Same data, same governance, whether it's an app, an agent, or someone on your team building their own workflow. 2 · Context is the whole game # There are three ways to give AI context about your business. Only one connects the dots. An AI answer is only as good as what the AI can see when it answers. Here's how the three common approaches stack up, in plain terms. Approach 1 ### Context Dump You paste everything into the prompt and hope. The AI skims, samples, and guesses. You pay for every page it reads, and the bill grows every time your data does. Approach 2 ### RAG The AI searches your documents or queries one system, then answers from what it found. Cheaper, and great for a single lookup. But it can't tell that two files describe the same job, crew, or asset. Approach 3 ### OAG (ontology-grounded) Every relevant fact is already connected, current, and sourced before the AI says a word. It asks one precise question of a connected map of your business, and its job shrinks to explaining the answer. 1 · The speed trap # A confident answer isn't the same as a correct answer. 22–94%of AI guesses are wrong When AI models lack an answer, they guess. The worst part? They don't tell you they're guessing. Stanford AI Index 2026, ch. 3; NIST AI 600-1 2 · Context is the whole game # How the three approaches stack up. ← Swipe to compare all approaches → | | Context Dump | RAG | OAG | | --- | --- | --- | --- | | Cross-system questions | ×Often wrong, confidently | —Right on one system, misses the join | ✓Built for them | | Cost per question | ×Highest | —Moderate | ✓Lowest | | As your data grows | ×Cost grows with it | —Cost stays capped, accuracy slips | ✓Cost stays flat | | Can you trace the answer? | ×No | —Partly | ✓Every fact has a source | | Safe to act on? | ×Not without redoing the work | —With heavy checking | ✓A person confirms, then it acts | 4 · Where Ontollo fits # Decide and act with confidence. Stream · Job 1931 True exposure and why the forecast moved Exposure Engine · Riverside PlazaMargin risk Ask below, then send — Ontollo drafts the moves; you keep the call. TN Tiffany McIntyreProject Manager · Job 19316:58 AM Exposure Engineworking Getting the answer Exposure Engineclose the loop6:59 AM `/task, draft the moves`ran Two actions clear most of the gap. The system does the typing; you keep the call. 1 Post #INV-61704 · $412,600 to #RiversidePlaza · CC 04-160 · #PO-3847 Three-way match: PO, L5 receiving log, invoice. Coding inherited from the PO. pilot 98% 2 Resolve duplicate: INV-61688 vs manual accrual JE-1840-031 Same vendor, amount and scope. Keep the invoice, reverse the accrual; count once. 94% Confirm & run Adjust Confirmed, running Reply in your own words… **@person** **#object** **/skill** 4 · Where Ontollo fits # A live view across your portfolio. Stream Company Projects Compliance Map TN New project Filter projects… Views Portfolio On track**2** At risk**3** Jobs *· 5* Riverside Plaza**71** Bayport Data Hall 2**76** North Bay Expansion**74** Med Center Pavilion**91** Katy Logistics DC**93** Search jobs, POs, people… TN Projects · Portfolio Every job’s path to its completion cost What’s posted, what’s committed but not yet visible to Dynamics, and where the forecast lands against budget. **2**On track · ≥80% confidence **3**At risk · 65 to 79% **0**Slipping · below 65% **$34.8M**Invisible to Dynamics Path to completion cost · *BUDGET tick · POSTED solid · COMMITTED ◇ · FORECAST EAC ●* Riverside Plaza Portland · Structural · $41.8M BUDGET EAC on budget, margin thinning: +$385K this week **71%**CONF Bayport Data Hall 2 Seattle · Structural · $33.4M BUDGET+$0.7M +$0.7M, retainage tail widening **76%**CONF North Bay Expansion Portland · Structural · $16.4M BUDGET+$0.5M +$0.5M, rebar rate drifting **74%**CONF Med Center Pavilion Seattle · Structural · $28.3M BUDGET On plan, under budget **91%**CONF 5 · Book a demo # See this on your own Agriculture data. Bring us one question like this one that your team answers by hand today. We'll show you what it looks like when the answer comes back connected, sourced, and ready for a person to act on. 6 · What your team will ask # The concerns are fair. Here's how each one gets handled. ### Is our data safe? Public AI tools can keep what you type, and some train on it. Company data stays in approved, access-controlled systems, with each person seeing only their slice. ### Will it give us wrong answers? On its own, AI predicts from patterns, so it can be confidently wrong. Every answer traces back to a source, and a person checks it before it counts. ### Our data is a mess. AI is only as good as the records it reads. Capture data once, where the work happens, and link it across systems instead of cleaning it forever. ### Will AI make decisions for us? It can only do what you give it permission to do. You set the boundaries. AI prepares the call. People make it. 6 · What your team will ask # Want the definitions? Four terms, in plain language ##### Ontology A shared rulebook for your business: what counts as an asset, a site, a crew, or a work order, and how they connect. Every system you run keeps its own private version. An ontology is the one everyone agrees on. [Tom Gruber, Knowledge Acquisition, 1993](https://tomgruber.org/writing/definition-of-ontology) ##### Knowledge graph Your real data mapped onto that rulebook: this machine, at this site, since this date. A web of connected facts a computer can walk, instead of searching blindly. [Hogan et al., ACM Computing Surveys, 2021](https://doi.org/10.1145/3447772) ##### RAG (retrieval-augmented generation) The AI searches documents or queries a system first, then answers from the handful of results that look most relevant. Good for one lookup in one place. [Lewis et al., NeurIPS, 2020](https://arxiv.org/abs/2005.11401) ##### OAG (ontology-augmented generation) The AI queries the knowledge graph directly and gets back connected, sourced facts. The facts come from the graph, not from the model's best guess, and the model's job is to explain them. [Related research on graph-grounded retrieval: Edge et al. (Microsoft Research), 2024](https://arxiv.org/abs/2404.16130) ← Previous Next → Construction Energy Health Care Operations Agriculture #### Navigate **1 · The speed trap**Point AI sees a slice **Copilot checks its workspace** **What Copilot could not see** **The cost of a confident guess** **2 · Context matters**Three approaches **Context Dump** **RAG** **OAG** **How they stack up** **3 · What it costs at scale**Cost calculator **4 · Where Ontollo fits**Disconnected systems **System of Reality** **Decide and act** **Portfolio in context** **5 · Book a demo**Book a demo **6 · FAQ**Frequently asked questions **Four definitions**