Frequently Asked Questions

Why isn’t AI alone enough for RevOps?
AI is excellent at reasoning, summarizing, and generating insights. But Claude, or any model, doesn’t know how your organization defines a qualified opportunity, calculates weighted pipeline, applies pricing policies, or resolves customer ownership.
Without that context, two people can ask the same question and get two different answers — because the model is guessing at rules that live in your business, not in its training data.
AIdeaBlocks captures how your business actually runs — its definitions, policies, workflows, and decision logic — so any AI, including Claude, applies the same rules every time.

Isn’t this just another RAG system?
No. RAG retrieves documents. AIdeaBlocks captures how those documents connect to what the business actually does. Instead of just handing back a pricing policy or a sales playbook, AIdeaBlocks knows how that policy relates to your entities, workflows, schemas, skills, application views, and decisions. The AI isn’t reading documentation — it’s operating against a structured model of your business.

How is this different from a Semantic Layer?
A semantic layer gets everyone to agree on what “Revenue” or “Customer” means. AIdeaBlocks builds on that — it captures not just shared definitions, but the rules that tell AI how the business should behave: policies, workflows, exceptions, and the logic behind a decision. A semantic layer defines the business. AIdeaBlocks lets AI run it.

How is this different from Knowledge Graphs?
Knowledge graphs describe how concepts relate to each other. AIdeaBlocks uses relationships too, but goes further — it captures logic that’s executable: the policies, workflows, and skills that actually drive a decision, not just the concepts behind it. A graph tells you how things connect. AIdeaBlocks tells AI what to do about it.

Why not just use prompts?
A prompt describes what you want in the moment. It doesn’t capture how your organization actually works. Without a shared model to draw from, prompts drift — every user writes their own version of the same business rule, and AI produces a slightly different answer for each of them. AIdeaBlocks separates durable business logic from one-off prompts, so it doesn’t have to be re-explained, and re-invented, every time.

What exactly does AIdeaBlocks capture?
The rules, context, and structure that today live in people’s heads, not in a system: business definitions, policies, business rules, workflows, skills, schemas, application views, intent interfaces, insights, alerts, and the relationships between all of the above. Together, these become the operational model AI executes against.

How does this actually run against my data?
Capturing the rules is half of it — AIdeaBlocks also builds Flows: the data prep, cleansing, and analytics pipelines that execute against your live data. Because those flows are built directly on the same policies and definitions your organization has captured, the logic governing the answer is the same logic governing how the data got there. That link is what makes the result deterministic — not just a well-reasoned guess, but the same calculation, run the same way, every time.

Do I still need a separate data prep or ETL tool?
No — and that’s one of the bigger differences. Most stacks force you to stitch together a semantic layer, a data prep/ETL tool, and a BI or dashboard layer as three separate systems, each with its own place for definitions to drift. AIdeaBlocks unifies the semantic layer and data prep/quality in one platform: profiling, cleansing, deduplication, and governance all run against the same artifacts that define your business rules. One platform instead of three stitched-together frameworks means fewer places for “Revenue” to mean something different depending on which tool computed it.

Can AIdeaBlocks act on its own, or only answer questions when asked?
Both. A flow doesn’t have to wait to be asked — it can run as an autonomous agent, watching for shifts in your data (a metric moving outside its normal range, a pipeline stalling, a policy exception showing up) and firing an alert the moment it happens. It’s the same governed logic used for on-demand answers, just running proactively — so you find out about a problem before it shows up in next week’s report, not after.

Does AIdeaBlocks replace Claude?
No. Claude remains the reasoning engine. AIdeaBlocks is what Claude reasons against — the score it performs, not the performer. Think of Claude as the intelligence, and AIdeaBlocks as your organization’s governed, versioned memory: recorded once, applied consistently every time it runs.

What problem does AIdeaBlocks solve?
Without a shared operational model: two users ask the same question and get different answers, pricing policies drift between teams, CRM definitions vary by department, prompts get longer and more brittle over time, and AI decisions become hard to trust or audit. AIdeaBlocks gives AI a governed model to work from — and you can show your work.

Is this only for RevOps?
No. RevOps is the clearest starting point because inconsistent AI shows up immediately — in forecasts, pricing, pipeline, and customer ownership. The same model extends to Marketing Operations, Customer Success, Finance, HR, Supply Chain, Procurement, and beyond.

What is an Operational Model for AI?
Every organization already has an operating model. People know the policies, the workflows, the exceptions, how decisions actually get made. AI doesn’t — until someone writes it down in a form AI can use. AIdeaBlocks captures that knowledge as governed, connected artifacts AI can understand and execute. The result isn’t just better AI. It’s AI that runs the way your enterprise runs.

Can I bring my own LLM?
Yes. AIdeaBlocks works with provided connections to OpenAI, Gemini/Vertex, and Anthropic — or you can bring your own API key and use the model of your choice. The governed artifacts — policies, flows, schemas — stay the same no matter which model is performing against them.

Is this a closed or open environment?
Open. AIdeaBlocks exposes MCP, so Claude and other AI tools can create and edit policies, flows, and other artifacts directly, not just read them. The knowledge repository can also be hydrated from, or accessed by, external tools — so the governance you build here extends beyond AIdeaBlocks instead of locking you into it.

How long does it take to get set up?

Most teams connect their first data sources and codify their first policy within a day. A full operational model — definitions, policies, workflows, and views for a business function — typically takes a week.

What about data security and where policies live?

AI is excellent at reasoning, summarizing, and generating insights. But Claude, or any model, doesn’t know how your organization defines a qualified opportunity, calculates weighted pipeline, applies pricing policies, or resolves customer ownership.
Without that context, two people can ask the same question and get two different answers — because the model is guessing at rules that live in your business, not in its training data.
AIdeaBlocks captures how your business actually runs — its definitions, policies, workflows, and decision logic — so any AI, including Claude, applies the same rules every time.

Can customer export their metadata from the system?

Yes. Metadata can be exported in a portable, human- and machine-readable format (JSON) and re-imported — into another AIdeaBlocks organization, for backup, or to preserve version history.