AI that actually knows your business.
The Business Intelligence Layer is AI trained on your own business documents. Plug in your files and anyone on the team can ask the business a question in plain English, and get an answer grounded in your real documents, not the open internet.
"Why does the AI know everything except how we work?"
It's the frustration that lands the moment a team starts using ChatGPT or Claude in earnest. Out of the box, these tools know everything about the internet and nothing about your business. They've never read your proposals, met your clients, or seen how you actually work. So every useful answer needs context spelled out first, and every answer needs double-checking. Meanwhile the knowledge that actually runs your business lives scattered across documents, drives, and a few people's heads, and it walks out of the door the moment someone leaves or takes leave.
The Business Intelligence Layer closes that gap. It's the foundation piece: AI grounded in your business rather than the open web.
What the Business Intelligence Layer is
In plain terms: you plug in your files, and the AI learns your business from them. After that, anyone on the team can ask the business a question in plain English and get an answer grounded in your real documents, not a generic answer from the open internet, and not something you have to re-explain the context for every time.
It's called a "layer" because it sits underneath the rest: the foundation the other Lume services get more useful on top of. Give AIOS this layer and its briefings and drafts are informed by how your business actually works; give a custom tool this layer and it answers with your knowledge, not the web's.
Who it's for, and when it isn't
This is for businesses whose real value lives in their documents and their people's know-how: consultancies, agencies, professional-services firms, anyone where "how we do it" is written down somewhere but hard to find fast. Three payoffs tend to matter most: new hires get productive from day one, knowledge doesn't walk out of the door when senior staff are on leave, and your best people's expertise becomes everyone's.
This isn't for you if you don't really have a body of documented knowledge yet. If the answers only live in people's heads and nothing's written down, there's little for the AI to ground itself in, and you'd get more from writing things down first. It's also not a fit if what you actually need is a task getting done rather than a question getting answered. That's AIOS or a custom build. And if your files are a genuine mess, we'll usually suggest a tidy-up first: the layer is only ever as good as what you feed it.
How it works, and where your data sits
The shape follows every Lume engagement: discovery to understand what knowledge matters and where it lives, design to agree what goes in and who should be able to ask it, a build you see progress on weekly, and an embed step where we train the team and stay involved.
On the question every owner asks first (where does our data live, and is it safe?), the answer is that your documents stay yours. Wherever possible we design it so your files stay inside systems you already control, rather than being copied somewhere new, and access is limited to the people you decide should have it. The AI models we build on don't learn from your data: your proposals, contracts, and internal files aren't used to train a public model. Before anything goes live we walk you through exactly where your data sits and who can reach it.
What you get
- A single place to ask your business questions: plain-English answers drawn from your own files, available to the whole team (or the parts of it you choose).
- Faster onboarding: new hires find answers themselves from day one instead of interrupting senior staff.
- Knowledge that stays put: the expertise doesn't leave when someone does, and it's reachable even when the person who wrote it is away.
- A foundation for the rest: the layer other Lume systems draw on to answer with your knowledge rather than the web's.
What people use it for
Common uses look like: asking "have we done work like this before, and what did we say?" across old proposals; getting a fast, correct answer on your own policies and processes without hunting through drives; pulling the relevant clauses or precedents out of past contracts; bringing a new starter up to speed on "how we do things here"; and settling the small factual questions that today bounce around the team on email. These are illustrative of how businesses use the layer. What yours answers depends on the documents you give it.
What it costs
As with everything we build, there's no single figure. The cost tracks the scope. What moves it: how much knowledge you're bringing in and what state it's in, how many people need access, and how clean and accessible your files already are (tidy, well-organised documents are quicker to work with than a sprawling, duplicated drive). A focused layer over one team's documents sits at the lower end; a business-wide knowledge base sits higher. We scope and agree it up front. The AI Audit is the low-commitment way to weigh it against your other priorities first.
Common questions
The general answer is: the documents your business actually runs on, things like proposals, contracts, policies, process notes, briefs, and reference material. The whole idea is that it learns your business from your files, so the more of your real working knowledge you can give it, the more useful it is. The practical detail of which exact file formats and storage locations we can connect to depends on how your business is set up, and we'll confirm what's in and out of scope with you during discovery rather than promise a blanket list here.
There isn't one fixed answer, and that's deliberate: how the layer is built (and where the searchable index physically sits) is decided per business, shaped by your setup and your security requirements rather than forced into a standard architecture. That design conversation happens early: if you have particular obligations (client confidentiality, sector rules, internal IT policies), tell us at the start and we build around them, not against them. And before anything goes live, we walk you through exactly where your data sits and who can reach it, so nothing about the arrangement is taken on trust. The homepage principle holds throughout: wherever possible, your files stay inside systems you already control.
Less than you'd fear, but some tidying pays off. The layer is only ever as good as what you feed it, so if your files are a genuine mess of duplicates and out-of-date versions, we'll usually suggest a light tidy-up of the worst of it before we build. You'd otherwise be grounding answers in stale documents. That said, part of what we do in discovery is work out what matters and what to leave out, so you don't need everything in perfect order first. We'll be honest about how much prep is worth doing versus what we can work with as-is.
The core safeguard is grounding: answers come from your own documents rather than the open internet, so the AI is working from your real material instead of guessing from the public web. That's the main thing that makes it more reliable than a general chatbot for questions about your business. As with any AI tool, it's a well-briefed assistant rather than an infallible oracle, and for anything consequential we'd always expect a person to stay in the loop. The value is in getting to a grounded answer far faster, not in removing human judgement.
Give your business a memory
If your best knowledge is locked in documents and a few people's heads, this is how you make it available to everyone, on demand. Tell us what kind of knowledge your business runs on, and we'll be in touch within two working days. Straight to Ben.
See if the Business Intelligence Layer fits →