We use the phrase “commercial intelligence” a lot at Keeve. It sounds obvious until you try to define it.
It isn’t business intelligence. It isn’t a dashboard of revenue, margin and pipeline. It isn’t having all your systems connected and it isn’t an AI assistant that can search across them.
So what is it?
What businesses already have
Most businesses have a lot of information. What they’ve sold, invoiced and delivered. Who their clients are and what their contracts say. CRM, project management, accounting, email, Slack and a growing pile of other tools.
What they don’t have is an understanding of what all of it means when you look at it together.
A client may have agreed a scope in a proposal, but the reason for that scope is buried in an email thread. A project may look profitable in the accounts while the team has quietly absorbed months of requests that never became change orders. A client may look like a good customer on revenue and margin while the conversation history tells a different story.
None of those facts is hard to find on its own. The value is in connecting them and that’s where it breaks down.
Why connecting your systems doesn’t solve it
The instinctive fix is to connect everything and point an AI assistant at it. That gets you better retrieval. It doesn’t get you understanding.
Take a client who’s bought the same kind of work several times. The CRM knows the client. The project system knows the projects. Finance knows the revenue and cost. The commercial insight, that this client consistently generates a specific kind of unscoped work or that this project type always needs more senior time than it’s priced for, isn’t sitting in any of those systems. It has to be inferred from hundreds of small decisions and requests scattered across conversations that were never designed to be connected.
An AI assistant layered on top of your existing tools can find the email where a client asked for another round of changes. That’s retrieval. It’s a different task from recognising that the client has done this on eight other projects, that the pattern reliably creates extra work elsewhere and that it should change how the next engagement is scoped and priced.
That’s understanding.
And this is the part that’s easy to miss: the problem isn’t that the information is hard to search. It’s that much of the commercially significant information was never recorded as a discrete fact anywhere.
People don’t log a CRM entry every time they spend ten minutes solving a client’s problem, make a judgement call or agree to an exception to keep a project moving. Each of those is reasonable in isolation. The pattern only becomes visible once you realise it’s happened twenty times and a system of record built to capture transactions was never designed to hold that.
A more capable AI assistant searching those same records more cleverly still doesn’t solve the underlying problem. The pattern isn’t sitting there waiting to be retrieved. It has to be constructed from the evidence.
From evidence to understanding
There’s a second problem underneath that one and it’s less about what the AI can see and more about what happens to what it learns.
Ask an LLM to analyse a collection of emails and it can produce a useful answer. Ask it the same question again later and it can analyse the underlying evidence again. But unless something persists between those interactions, nothing accumulates between one answer and the next.
That’s an important distinction.
A persistent commercial model provides the structure between the evidence and the answer. Instead of re-deriving everything from raw text each time, it can hold an established understanding of the client, the project and the patterns that have emerged: this client regularly generates this type of unscoped work; this project type consistently requires additional senior involvement; this concession tends to have this consequence.
That understanding can be built from evidence, refined as new evidence arrives and carried forward into the next decision.
Ask the same question again and you’re not asking the system to rediscover the answer from scratch. Ask a follow-up and it can build on what it already knows. More importantly, the conclusion can be traced back to the evidence that supports it.
That’s the difference between an AI that can produce an answer and a commercial intelligence system that can maintain an understanding of the business.
What commercial intelligence actually has to do
Commercial intelligence has to work backwards from evidence that was never designed to be connected, recognising patterns, relating events across systems and understanding consequences.
A client request is one thing. The work it generates elsewhere is another. An exception made to keep a deal moving is one thing. What it does to margin or the next negotiation is another.
That’s the difference between reporting and intelligence. Reporting tells you margin fell. Commercial intelligence tells you why, whether it’s a one-off or a pattern and what to do differently next time, in pricing, scoping, resourcing or how you handle that client.
The value was never in knowing that something happened. It’s in understanding the pattern underneath it, what it means and carrying that understanding into the next decision.
That’s what we mean by commercial intelligence at Keeve.
Not more information. Not another dashboard. Not simply a smarter way to ask questions of the data you already have.
It’s an understanding of your commercial reality built from what actually happens, including everything that never made it into the record.
Most businesses already have systems of record.
What they’ve never had is something that understands the record, the conversations around it and the patterns underneath them, then carries that understanding forward into the next commercial decision.