Skip to content

02.Services

AI Automation & Agents

Agents are a decision before they are a build. I start with which work an agent should take, which stays with a person, and what a wrong answer would cost. Then I build it: agents and multi-agent workflows that run inside the systems you already have, with every action recorded.

~/thelogicatelier · services/ai-automation-agents/

AI Automation & Agents

Agents are a decision before they are a build. I start with which work an agent should take, which stays with a person, and what a wrong answer would cost. Then I build it: agents and multi-agent workflows that run inside the systems you already have, with every action recorded.

An agent engagement starts with a boundary written down: which tasks an agent may complete on its own, which it prepares for a person to approve, and which it may not touch. Then I build inside the systems you already run, with high-risk actions held for human sign-off, every action recorded, and the running cost settled before the build.

What an agent is allowed to do.

My first piece of work is a boundary, written down: which tasks an agent may complete on its own, which it may prepare for a person to approve, and which it may not touch. The question behind it is what a wrong answer would cost, and that answer differs for a draft document and for a production change.

High-risk actions are held for verified human sign-off rather than trusted to a model, and every action, tool call and prompt is recorded, so a reviewer can see what happened and in what order.

Built into the systems already running.

Automation that sits beside a business adds a system to maintain. Automation wired into what already runs removes work instead. So I integrate with the tools you have, through Model Context Protocol integrations where other teams need to reach the same capability from their own software.

Evaluation and running cost are part of the build. The measures that show an agent still behaves as agreed are the same ones that show what it costs to run, and both are in place before I hand the work over.

On the record.

On the internal AI workflow platform for EY, agentic workflows automate threat modelling, compliance, retrieval-grounded documentation, planning, code generation, security scanning and reviews. A trust-tier model holds high-risk agent actions for verified human sign-off, and an audit trail records every agent action, tool call and prompt.

For Walgreens Boots Alliance, delivered through EY, agents took on the quality-assurance load: a SonarQube remediation agent and an automated code review agent, with adoption across client teams driven by exposing the framework through a Model Context Protocol server. The testing framework that followed has its specifications generated automatically through that server.

What the work covers.

  • Agent and multi-agent workflow design
  • Automation inside the systems already in place
  • Model Context Protocol (MCP) integrations
  • Human sign-off for high-risk actions
  • An audit trail of every action, tool call and prompt
  • Agent evaluation and running-cost control

Questions.

Which work should an agent take?

That is the first thing I settle with you, in writing: which tasks an agent may complete on its own, which it prepares for a person to approve, and which it may not touch. The question behind each one is what a wrong answer would cost, and that answer differs for a draft document and for a production change.

How are high-risk actions controlled?

They are held for verified human sign-off rather than trusted to a model. That is the trust-tier model already in production on the internal AI workflow platform I built for EY, where an audit trail records every agent action, tool call and prompt.

Do agents work with the systems I already run?

Yes, and that is the point. Automation wired into what already runs removes work, while automation sitting beside it adds a system to maintain. Where other teams need the same capability from their own software, I expose it through Model Context Protocol integrations.

What does it cost to run, and how do I know it still behaves?

Evaluation and running cost are part of the build, not a later optimisation. The measures that show an agent still behaves as agreed are the same ones that show what it costs to run, and both are in place before I hand the work over.

Start with the free assessment.

Thirty minutes, free, on a video call. You bring the challenge, I tell you where I would start and what the work would take. No obligation.

The assessment is free. Anything that follows is quoted in writing before it starts.

Book the free assessment

vi.Contact

06 · contact

Talk to me.

Talk to me.

Production AI and software architecture. Based in Dubai, working with teams in the Emirates and Ireland.

I work Dubai hours, Monday to Friday, the same week you do. Dubai is three hours ahead of Dublin in summer and four in winter, so my afternoon is your morning.

I will never sell your details, for any reason, and I never send spam.

How I use your enquiry
  • Your enquiry comes to me, Abhijit Chatterjee Chattopadhyay of The Logic Atelier. The form is handled by Formspree, which passes it to my inbox and keeps a copy in my Formspree account.
  • I use it to read your enquiry and reply. If no work follows, I delete it within 12 months, and sooner whenever you ask.
  • If Google Analytics is running on your visit, it is told that an enquiry was sent, and nothing you typed.
  • To see or delete what I hold about you, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.