AI Assistant
One assistant for the whole company. Under your control.
One AI assistant for everyone in your company, from reception to the board. It knows your documents, your numbers and your agreements, and every answer shows where it came from.
In plain terms: the AI chat your people already want, but it knows your company and your IT team decides what it may see. One fixed price for the whole organisation, no licence per user. Hosted in Europe.
What it is
Most AI tools see a question, rarely the whole company. The useful answers sit in your documents, in your data, in what your people remember and in how it all connects. And the real work does not stop when the conversation closes. This assistant is built for all four.
Knows your company
Your documents, with the source, and who and what each one concerns.
Talks to your numbers
Live figures in plain language, with the definition behind each one.
Remembers and learns
What your team decides and learns stays for the next conversation.
Takes work over
Describe a recurring task, approve the steps, and it runs on its own.
One question, two sources
What that looks like on an ordinary Tuesday. One question, answered from a document and from a system at the same time, and the answer says which is which. Nobody had to know which system holds what.
What people do with it
Six questions real people ask, and what comes back. Each one names the parts that did the work.
“What did we propose to this customer last time, and what did they push back on?”
The previous proposal, the meeting notes and the lesson a colleague recorded afterwards, with links to each.
“Which products lost margin last quarter?”
The figures, the definition of margin your company uses, and a link to the report they come from. Only for the people allowed to see them.
“A customer says the amount on their invoice is wrong. What do I do?”
The steps your own policy prescribes, in order, with the paragraph they come from. Not a colleague interrupted for the ninth time this month.
“Remind customers of open invoices every Monday.”
You described the flow once and approved the steps. Every Monday it drafts the reminders, sends what you allowed, and reports what is still open.
“Set up everything for the new colleague who starts on the first.”
Accounts requested, welcome mail drafted, the onboarding checklist turned into tasks for the right people, and a nudge when one stalls.
“Prepare my meeting with this supplier tomorrow.”
The last mails, the open actions, the numbers on the contract, and what went wrong the last time. Read as you, so you only see what you may.
And the ordinary stuff all day long, connected to the mail, calendar and files your team already lives in: draft the mail, find the slot, summarise the report, turn the brief into a deck.
From a conversation to a flow that runs
The part that makes it more than a better search box. Someone explains how a process works, the assistant turns it into steps, a human approves, and from then on it runs. No low-code canvas, no developer, no waiting.
- 1
Describe it
You tell the assistant how the process runs. It asks until the steps are clear.
- 2
See the blocks
The steps appear as building blocks, so you see exactly what will happen.
- 3
Approve
You sign it off. For a company process, the process owner does.
- 4
It runs
With versions, a way back, and a trail of who changed what.
- 5
Everyone can see it
The flow sits in a shared overview, so nobody builds it a second time.
Personal
Prepare my meetings, sort my mail.
No approval needed. It is your own time.
Team
Weekly follow-up of candidates.
The team lead approves. The team can see it.
Company
Payment reminders, onboarding a new colleague.
The process owner approves, with versions. Everyone can see it.
Where a job has rules, a specialist sits behind the general assistant: accounting gets an agent that checks bookings against the documents behind them, marketing one that checks campaigns the way campaigns get approved here. Nobody picks from a menu. People ask, and the assistant hands it to the one built for that job.
Under the hood, for whoever signs it off
Most people never think about this part. They ask a question and get an answer with a source. If your job is to check how that works, these are the four pieces.
- 1
The knowledge engine
Reads your documents before it indexes them, cites the paragraph it used, and returns nothing rather than leaking when two systems disagree about access.
How the engine works - 2
The relationship map
Who worked on which project, which project served which customer, which document explains which system. The documents stay the proof. The map finds the right one.
- 3
The data layer
Connects to the data platform you already run and answers in your own business definitions, so margin means what it means in your company. It reads as the person asking, so nobody sees a figure they could not open already.
Why the definitions come first - 4
The Brain
The memory. What you and your team learn, decide and agree on, kept for the next conversation. Not locked to our assistant: Claude, ChatGPT and the coding assistants read the same memory.
What the Brain is
One question, “who did something like this for a customer like them, and what did we learn?”, lights up one path. In your company the map holds thousands of these links, rebuilt from your sources, never typed in by hand.
Runs where you decide
Every customer gets a dedicated environment, the models run in your own cloud account, and the rules stay yours.
Managed by us, in Europe
- Your own isolated environment on our European servers
- The AI models run in your own cloud account
- Updates, hosting and support handled by us
- Nothing shared with another company
Or in your own environment
- The whole platform on Azure, AWS, Google Cloud or your own servers
- For larger organisations with a mature IT team
- Same product, your infrastructure
- Updates in agreed windows
Either way: pick the model that fits each department and switch any time, Claude, OpenAI or open-source through one gateway. Single sign-on through your own identity provider, audit on everything. GDPR compliant and EU AI Act ready.
Governed by design
The point of one platform is that someone is actually in charge of it. IT gets the controls and the audit trail, and your people just get an assistant that works.
How a rollout goes
Measured in weeks, not quarters. The proven route starts small on real work and grows department by department.
Your environment, ready in days
Managed by us in Europe as the standard, or in your own environment if you prefer. People sign in with the account they already have.
Connect the first knowledge base
The knowledge engine indexes the documents one department actually uses. Selective on purpose: index what people use, not everything you store.
Pilot on real work
A test group uses it on their actual tasks. Their questions and complaints tune the setup better than any workshop.
Grow team by team
Each new department gets its own space, knowledge and rules. The flows people build become shared, and use cases surface as you go.
Two ways to get it
Same building blocks, a different amount of us. Most companies start with the platform as it is. Some need it built into the way their business actually runs.
The platform, as it is
Your own environment in days, one fixed monthly price for the whole organisation, nothing to build. Screenshots, the full feature list, security detail and current pricing live on the product site.
Visit aiassistant.plainsight.proBuilt into how your company works
When the answers sit in systems only your company has (a MES, an ERP, the planning tool someone built in 2009), our team builds the assistant into them: your sources, your flows, your rules. A project with our experts, not a subscription.
Talk to our teamFrequently asked questions
What can people actually do with it?
The everyday work: ask questions that get cited answers from company documents, draft and triage mail, analyse data and build charts, and turn briefs into presentations. One governed place to ask, instead of a different subscription per team.
Is it more than a chatbot?
Yes. It knows your company: your documents, and how people, projects and systems connect. It talks to your numbers, with the business definition behind each figure. It remembers what your team learns. And it takes recurring work over: you describe the task, approve the steps, and it runs and follows up. Each part is useful on its own.
Where does our data go?
Into an environment that is yours alone, and nowhere else. As the standard we run it for you on our own European servers, isolated per customer, with the AI models in your own cloud account. Larger organisations with their own IT team can run the whole platform privately instead, on Azure, AWS, Google Cloud or their own servers. Either way your data stays in Europe, is never used to train models, and people sign in with your existing identity provider.
How does it know our documents?
Through the knowledge engine we built ourselves, which is part of the assistant and included in the platform price. Each department gets its own knowledge base, answers cite their source, and nobody sees a document their permissions would not already open. Behind it sits a map of how people, projects, customers and systems relate, so the assistant can answer how a previous project was handled, not only what a policy says.
Can we control who sees what?
Yes, and it is the default rather than a setting. Departments are isolated from each other with their own knowledge bases and rules, permissions are enforced per document, and an admin dashboard shows usage, costs, models and users.
What happens when the assistant runs a task on its own?
Nothing runs that a human did not approve first. You describe the task, the steps appear as blocks, and the owner of that process signs them off. From then on the flow runs with versions, a way back, and a trail of who changed what. Personal flows need no approval; team and company flows do.
Why not just use what is in our Microsoft licence?
We work with Microsoft rather than against it: the assistant plugs into the Microsoft 365 your team lives in, and runs in an environment that is yours alone. What we add is model-vendor neutrality (pick the model that fits each department, switch any time), knowledge bases with per-document permissions, and platform pricing instead of per-seat licences. We pick what fits, we are not tied to any vendor.
What does it cost?
One fixed monthly amount for the whole organisation, sized by company size, plus a one-off onboarding. No licence per user, so adding a colleague costs nothing extra and nobody has to decide who gets access. Because the models run in your own cloud account, the admin dashboard shows the estimated cost per model and every user gets a sensible usage limit, so the bill does not surprise you. Current numbers live on the product site's pricing page.
How fast can a team start?
Setting up your environment plus connecting the first knowledge base is measured in weeks, not quarters. The proven route: start with one or two departments on real work, then scale team by team as use cases surface.
Ready to give everyone in your company the same assistant?
Tell us how your teams use AI today, official or otherwise. We'll show you the assistant on real work: your documents, your numbers, your agreements, in an environment that is yours.
Ask for a demo