SVENNISAIAn AI agent on Claude takes over one process in your company.
Orders, quotes or support requests: the agent handles them in your systems, and a colleague approves. On day 30, the result sits beside the signed targets.
from 16,800 GBP
All prices are in GBP and exclude VAT.
Book a 30-minute call · with a Claude Certified Architect from our team, who builds the agents.
What you get
Pilot A is for a company with a process that takes hours every day. The agent runs on the API: a program calls Claude directly, with nobody at the keyboard.
- 01An AI agent on Claude working on one process: orders, quotes or support requests.
- 02The agent works in the systems you already use, on your company's own account, with your keys.
- 03The approval step: at first, a colleague confirms every action. You decide how long it stays.
- 04The day-30 review: the result beside the signed targets, and what Claude cost.
- 05The log: every action of the agent recorded on your account, open to check at any time.
One whole process, connected to your systems
The base price starts from 16,800 GBP and covers everything the process needs to run end to end. The existing connectors are in the price, and for a system we have not connected before we build an MCP server.
- 1 processend to endFrom the request that comes in to the entry in your system, with a colleague who approves.
- Allexisting connectorsMicrosoft 365, Google Workspace, Zoho, Xero: as many as the process needs.
- 1new MCP serverFor a system with an API we have not connected before, Sage or FreeAgent for example, with up to 8 read and write operations.
- Allthe process's skillsUsually 3 to 5, each with its own test cases.
- Englishplus a language on requestFor customers or suppliers who write in another language.
- 1channelEmail, the website form, WhatsApp or Microsoft Teams.
- 100test casesFrom your real work, checked before go-live.
- Yourseverything we buildThe agent, the skills and the tests stay on your company's account.
- A second new MCP server
- A system with no public API: we show you the options
- A second channel
- A second process: Department Activation or a new pilot
You get the final price in the written offer, before you sign. It does not change along the way without your agreement.
The same Monday morning, twice: Claude on a seat, then an agent
On a seat, Claude helps when someone asks. The agent takes the whole process, on its own, at every new request.
Monday, 08:00, a 25-person distributor in Leeds.
Twenty-three emails came in over the weekend. An illustration, built from real projects; the names and figures are not a client's. Each window plays as it comes into view.
What you gain, in this example: About two hours of one person's day, every day. Every weekend enquiry gets a reply within minutes. The week starts on Monday with four approvals, not twenty-three emails.
Worth knowing: Claude on a seat does each of these steps when a colleague asks, and since February it can also run a scheduled task. What it cannot do is read each email as it arrives, work in the company's name or follow rules tested on your data, with a log: that is the job of an AI agent on Claude. If the task comes up a few times a day, a seat is enough, and the Workshop says so.
How it unfolds, step by step
Pick a step: what it covers, what we need from you, what you keep.
Week one: the targets, the account, the test set
- At kickoff, the first working session, we sign the targets: the figures we measure on day 30, and where they stand today.
- Whoever signs off the budget signs them, usually the managing director. We start after the signature, because on day 30 we compare the result with them.
- You gather the test set: up to 100 real cases from recent months, each with the right answer from your team.
- You open the account in your company's name: with Anthropic, or with AWS in London. We help you open it; it is never on our account.
- Together we set the approval step: the point in the process where a colleague confirms.
You bring: The process you chose, the colleague who does it today, and whoever signs off the budget, to sign the targets. Real cases from recent months with the right answer, and access to the systems we connect.
The four steps
You buy the steps one at a time. At the end of each you hold something concrete, and you decide whether to go further. The pilot has two options, A and B; both are step 2.
- Step 1Claude Readiness WorkshopThe company that wants to know what is worth automating first, before it pays for a pilot.2,500 GBP
- Step 2 · You are hereClaude Pilot A: an AI agent on ClaudeThe company with a process that takes hours every day: orders, quote requests or support requests.from 16,800 GBP
- Step 2Claude Pilot B: Claude Team for your teamThe company that wants its team on Claude every day, on its own accounts, with rules and controlled access.from 12,500 GBP
- Step 3Department ActivationThe company that has been through the pilot and wants Claude in every department, on its real tasks.from 5,600 GBP a department
- Step 4Managed AI OperationsAn assistant or an agent we built, which you want kept in good order.from 550 GBP a month
All prices. Claude seats are billed to you directly by Anthropic, in pounds; an agent's model usage by Amazon Web Services, in US dollars, on use and not capped unless you set a budget.
Frequently asked questions
What does the human approval step mean?
The point where a colleague confirms before an action leaves: the quote before it reaches the customer, the bill before the ledger. The agent prepares, the colleague approves or corrects, and sensitive cases always go to a person. The step stays as long as you want; the review shows how often a correction was needed.
Do we need Claude seats for an AI agent?
No: an AI agent runs on the API, on your company's account, and you pay the model usage. At a few thousand requests a month, that is usually tens of pounds. Seats, the per-person licences, are for Claude Team or Enterprise, which is Pilot B.
Pilot A or Pilot B?
Pilot A when you have one process to hand over end to end, day and night, with nobody at the keyboard. Pilot B when you want the team on Claude Team, under rules, working with Claude themselves. The Workshop tells you which; if you skipped it, we say so on the first call.
What if the agent misses the targets on day 30?
At the start we make sure the agreed targets are right and reachable in production; we do not accept requirements that cannot be met. The review gives the figures against the targets signed at kickoff. You decide: extend, adjust the scope or stop, and what we built stays on your account.
Where does the data go?
To your account: the agent runs in your company's own AWS account in London by default, or in your own Anthropic organisation, with your keys. Which models run in the UK and which are processed within Europe, we write in the offer, with the regions.
Our ledger is Sage or FreeAgent. Can the agent work with it?
Yes: neither has a ready-made Claude connector today, so we build an MCP server on its API. The base price covers up to 8 read and write operations, and the offer states the scope before you sign.
The other offers: Claude Readiness Workshop · Claude Pilot B: Claude Team for your team · Department Activation · Managed AI Operations · Claude Team Adoption · Claude Relaunch · AI service desk assistant · AI customer assistant.
Last reviewed: 18 October 2026.
Book a 30-minute call
Tell us what you want to solve in your company. At the end you know whether it is worth it and what the next step is; the prices are the ones on this page.
Prefer to call? +44 7588 772736 · Not ready to talk? See it on your business first.
- 1Time
- 2Details
- 3Booked