The daily file hunt.
Someone spends a couple of hours every day finding attachments in email and stitching them into one big file. An agent does it every morning before they log in.
Service 03
From one person using Copilot to every team using AI. With the numbers to prove it.
We map your processes, show where AI pays off, run the pilot, build agents for each department and scale it across the company — on your own Microsoft or Google environment, with a budget you control and an effect you can measure.
Executive summary
Most companies are somewhere between “the board wants an AI plan” and “we bought licences and three people use them”.
The questions are the same everywhere. Where do we start? In what order? How do we run a pilot that shows results quickly? What is possible, what isn't, and what will it cost? Who has access to which data now that Copilot or Claude is switched on? Why is the bill growing while usage isn't? And how do we know any of this is working?
We answer them with a method, not a demo.
We break your company's processes into small steps and measure what each one costs today — in time and money. That shows exactly where AI saves the most. We set targets for one year and three years, run the first pilots, build small agents for each department, and roll them out with training, access rules and cost control. Then we keep watching the bottlenecks and fixing them.
You always know how much you invest in AI, where it goes, and what effect you expect.
When this fits
This is for you if:
- Leadership asked for AI, and nobody knows where to start, in what order, or what it will cost.
- You have Copilot, Claude or ChatGPT licences — and only a handful of people really use them.
- Copilot or Claude is switched on, and nobody is sure who can see which data.
- AI infrastructure and usage costs keep growing, and nobody can say what they buy.
- You need to show the effect of AI in numbers, not anecdotes.
- Your teams spend hours on work an agent could do: collecting files, merging spreadsheets, reading documents.
This is probably not for you if:
- You're about to retire. Then it's your successor's AI budget.
What we do
AI readiness and roadmap
- In practice
- Where AI fits in your business, in what order, what it will cost, what to expect
Process map and effect model
- In practice
- Every process broken into steps, current cost and time per step, targets for 1 and 3 years
Pilots with fast delivery
- In practice
- One or two use cases live in weeks, with results you can show the board
Department agents
- In practice
- Small agents that take repetitive work off each team
Rollout and adoption
- In practice
- Training, champions in each team, from “one person uses it” to “everyone does”
Access and security
- In practice
- Who sees what: permissions, data sources and policies mapped and fixed
Cost control
- In practice
- Licences, tokens, infrastructure and connectors — what you pay for and what you actually use
Your own AI workspace
- In practice
- For companies that want to run AI independently: an enterprise workspace like Langdock, set up and governed in-house
| What | In practice |
|---|---|
| AI readiness and roadmap | Where AI fits in your business, in what order, what it will cost, what to expect |
| Process map and effect model | Every process broken into steps, current cost and time per step, targets for 1 and 3 years |
| Pilots with fast delivery | One or two use cases live in weeks, with results you can show the board |
| Department agents | Small agents that take repetitive work off each team |
| Rollout and adoption | Training, champions in each team, from “one person uses it” to “everyone does” |
| Access and security | Who sees what: permissions, data sources and policies mapped and fixed |
| Cost control | Licences, tokens, infrastructure and connectors — what you pay for and what you actually use |
| Your own AI workspace | For companies that want to run AI independently: an enterprise workspace like Langdock, set up and governed in-house |
Every process has a price. We break your processes into small steps — typically around ten process groups with around ten sub-processes each — and record what each step costs in time and money today, and what it should cost in a year and in three. That map becomes your AI plan, your budget and your scorecard. Back-office processes are usually where it pays off first.
Examples of department agents
Ten spreadsheets, one report.
A team consolidates data from ten Excel files into a single report every week. An agent pulls, checks and merges them, and flags what doesn't add up.
From request for quote to proposal.
A construction company receives tender requests with ten or more attached files. An agent reads the documents, pulls out the scope, prepares a preliminary estimate and sends it into an approval workflow — so the estimator starts from a draft, not a blank page.
How it works
- 1Audit — 2–4 weeks. We map your processes or a chosen department, measure the effort in each step and find where AI saves the most.
- 2Choose the use cases. A short list with expected effect, cost and risk for each. You pick.
- 3Pilot. One or two use cases built and used by real people, with the effect measured against the baseline.
- 4Scale. Agents for more teams, training, access rules, cost control — rolled out department by department.
- 5Run and improve. A standing process: the process map is updated, bottlenecks are monitored and fixed, targets are tracked.
What we need from you: a sponsor in leadership, one owner per department, and access to the systems the agents will work with.
What you get
- A process map with prices. Every process and sub-process, current effort and cost, targets for one and three years.
- An AI roadmap and budget. What to do first, what it costs, what effect to expect.
- Working pilots — not slides.
- Agents in each department that remove repetitive work.
- Clear access and security. Who sees what, documented and enforced.
- Cost under control. Licences and usage matched to real value.
- Adoption. People trained, champions in each team, usage you can see.
Example
A large UK manufacturer — Microsoft AI from the data up.
The company had Microsoft 365 and Copilot licences and almost no use of them. We built the data infrastructure and ETL pipelines, a context layer over company data so AI answers from the company's own knowledge, and the first AI agents. Then we trained the business teams on Copilot. The client's team now runs it; we stay on for new agents.
Tools and platforms
- Microsoft Copilot · AzureFor companies on Microsoft 365: Copilot, agents and data in your own tenant
- Anthropic / ClaudeAssistants and agents that work with your documents and systems, connected via MCP
- OpenAI / ChatGPT EnterpriseEnterprise assistants and AI-enabled analytics
- Google Cloud / GeminiFor companies on the Google stack
- Langdock (or similar)An enterprise AI workspace when you want to run AI independently, across models
- Microsoft Fabric · Snowflake · DatabricksThe data platform the AI layer stands on
Formats
- Project. Audit and roadmap, or a pilot with a defined result.
- Outsourced function. We run your AI programme month by month: new agents, monitoring, cost and access control.
- Team extension. AI developers, AI systems architects or forward deployed engineers join your team.
Pricing on request.
Security and data
Your data stays inside your company. We build on your own Microsoft or Google environment, or a workspace you control, with access mapped to roles. Before rollout we check who can see what — and fix it.
FAQ
Where do we start if we have nothing yet?
With the audit. In a few weeks you know where AI saves the most, in what order to do it and what it will cost.
We already have Copilot. Why is nobody using it?
Usually three reasons: it can't see the right data, people don't know what to ask it, and nobody owns adoption in each team. We fix all three.
Is our data safe?
It stays in your environment, under your security rules. We also check and fix who can access what before anything is rolled out.
How do you measure the effect?
Against the process map: the time and cost of each step before and after. Not “people feel faster” — hours and pounds.
Which AI should we choose: Copilot, Claude or ChatGPT?
It depends on your stack and your use cases, and it's often more than one. We'll recommend based on the process map, not on a partnership.
Can we run AI ourselves afterwards?
Yes — that's the goal. For companies that want full independence we set up an enterprise AI workspace like Langdock and train your team to run it.
Why are our AI costs growing?
Usually unused licences, expensive models doing simple work, and connectors nobody needs. We map usage against value and cut what doesn't pay.
Where would AI save your team the most time?
Discuss your AI plansWe'll start with your processes, not with a tool.