Seats, not work
Buying ChatGPT or Claude seats makes individuals faster. It changes no process, so nothing reaches the P&L.
Most AI transformation programmes end with a roadmap and a few pilots. Ours end with AI agents running in your departments, assisting people or taking over defined parts of their work, on a knowledge layer and an environment in theagency47's Cloudflare account or yours. Every phase maps to a package with a published, fixed price.
New to the term? Read our guide, What is AI transformation, then come back here for how we run it.
At theagency47 it is not a separate product. It is the same agents, company brain and OS47 workspace we build for every client, sequenced into one programme for companies with 25 to 250 staff and for organisations above 250. Enterprise AI transformation and a mid-size roll-out follow the same four phases; only the starting point and the contract differ.
Buying ChatGPT or Claude seats makes individuals faster. It changes no process, so nothing reaches the P&L.
A strategy deck names twenty use cases and builds none of them. The team that wrote it moves on. (AI agent vs AI consultant explains the difference.)
The first agent works, but it was built on its own knowledge and its own rules, so the second one starts from zero.
Nobody decides which work an agent may assist with or take over, so every decision goes to committee. See who owns the decision, CEO or CIO.
We built the programme around those four failure points: work instead of seats, builders instead of decks, one shared foundation instead of isolated pilots, and a named owner for every agent.
The phases are company-level. Inside each build, our five-phase delivery method still applies. The month-by-month version is in the AI transformation roadmap.
We put hours against the work in each department, decide task by task where agents assist and where they take over, and agree the rules: what an agent may do alone, what needs approval, who owns it. You get a phased plan with a fixed price per phase. For 25–250 staff this is the Transformation Assessment.
One department goes first. We build its agents, the first stage of your company brain and the acceptance tests every agent must pass. Success is measured in hours and outcomes, not demos.
Further departments join on the same knowledge and the same environment, with permissions per audience, a log per agent and a named delivery lead.
Health checks, knowledge updates, new agents on a set cadence and a quarterly review of what to hand to agents next.
| Phase | What you get | Mid-size (25–250 staff) | Large (250+ staff) |
|---|---|---|---|
| Map | Process map with hours, assist or take-over decision per task, rules, phased plan | Transformation Assessment, €3,000, 4 weeks, or the AI Audit, €297, as a first conversation | Scale package scoping, €5,000, 2 weeks, credited 100% within 60 days |
| Prove | One department live, company brain stage 1, eval suite | Workforce Starter €7,500 or Workforce Pro €15,000, plus Second Brain €4,500 (€1,500 off when bundled) | Workforce Pro €15,000 as the first milestone, or straight into the Scale package |
| Scale | More departments, one shared brain with permissions, one environment | Workforce Pro €15,000 per further department; Run package €14,500 | Scale package from €25,000 (8–15+ agents, multi-department, 90+ days, SLA) |
| Run phase | Monitoring, knowledge upkeep, new agents | Grow €1,200 / month or Scale retainer €2,500 / month | Embedded €5,000+ / month |
Starting in sales? The AI Sales Agent (from €4,500 build + €750 / month) can be your Prove phase. Accounting, legal, healthcare, e-commerce and SEO/GEO firms can start from a vertical pack (€7,500 to €12,500). For customer service, Orion is an example of a ready agent that can go first, priced per resolution (€0.05 to €0.15). Model and Cloudflare usage are billed to you directly, at cost. All prices are on /pricing/.
A transformation is not one decision. It is fifty small ones. For every role in scope we ask four questions: how rule-based is the work, what does an error cost, who approves the output, and how often does the work arrive. The answers place each task on one of our three organisational levels (executive, operational, task) and set how far the agent goes. The full test, with ten worked roles, is in assist or replace.
| Pattern | What the agent does | Typical work |
|---|---|---|
| Assist | Prepares, drafts and summarises; a person decides | Board briefings, account reviews, contract first reads |
| Co-work | Handles the routine share end to end and escalates the rest with context | Tier-1 support, sales development, finance operations |
| Take over the task | Owns the task; a person reviews exceptions and samples | Data entry, inbox triage, document extraction, recurring reports |
How the people around those tasks adopt the change is its own discipline; see AI change management.
You have departments but no AI team. Start with the department that loses the most hours, usually sales operations, customer service or finance. One Workforce Pro build puts five coordinated agents into that department in 60 days, with a 14-day money-back guarantee. Want the hours mapped first? The Transformation Assessment (€3,000, four weeks) gives you a phased plan with a fixed price per phase. Your company brain starts as plain files your team can read and grows when the second department joins. Why one department beats a strategy deck: AI for mid-size companies.
You have procurement, security review and departments with their own systems. Start with paid scoping: in two weeks you get an agent inventory with one-page specs, a compliance posture document and a milestone-based, fixed-price proposal. The Scale package then runs 90 days or more under an SLA contract, with a dedicated delivery lead and a monthly steering committee with your executive sponsor.
Every department's agents read from one company brain and run in one environment. The AI Company Brain starts as plain files your team owns (the Second Brain with Obsidian), adds a search index when several agents share it, and adds permissions, memory and logs per audience when more departments and customers join. In OS47, your team sees and approves what every agent does, in theagency47's Cloudflare account or your own. Nothing is migrated as you grow.
Read how we keep agents under control on our Trust page and in our AI agent security guide. Who decides what each agent may do, company-wide, is set out in our AI governance framework.
The prompts, the configurations, the knowledge files, the search index and the eval suites. You can run everything yourself or move to another partner. We do not lock you in.
The first department is live in 30 to 60 days. A multi-department roll-out runs 90 days or more. Most clients plan in 12-month horizons with a quarterly review. Our AI transformation roadmap shows the four phases month by month.
Each phase maps to a published package: Transformation Assessment €3,000 or Scale package scoping €5,000, a first department from €4,500 (AI Sales Agent) or €7,500 (Workforce Starter), multi-department work from €25,000, retainers from €1,200 a month. Model and cloud usage are billed to you at cost. For a full year, line by line, see AI transformation cost.
Agents take over tasks, not whole people. We decide task by task where an agent assists, works alongside a person or takes the task over, and the people involved move to approval, exceptions and the work agents cannot do. See assist or replace, role by role.
No. Agents read the documents you already have, written once into your company brain, and query live systems where they are, read-only wherever possible. The difference is explained in AI transformation vs digital transformation.
Yes. For larger companies we provide a pre-filled vendor risk questionnaire, data-flow diagrams per agent, a sub-processor inventory, an access-control matrix and a controls mapping, and we sign MSA, SOW, DPA and SLA.
A plain definition for leadership teams, and what changes first.
Four phases from first department to whole company, with gates and owners.
Six differences, and whether you need one before the other.
Owners, autonomy levels, approval rules and logs, with a one-page agent charter.
How teams adopt agents and keep trusting them.
A 12-month budget for mid-size and large companies, line by line.
What a CAIO actually owns, why most mid-size firms don't need one full-time, and three ways to fill the gap.
Talk to our AI agent
Ask about our services, prices or how a project runs, in Greek or English. It answers out loud.
Uses your microphone. Ends automatically after 3 minutes.
Test version; the AI can make mistakes. Calls are recorded and the transcript is sent to our team; audio is processed by xAI. Privacy policy