1. Home
  2. Services

AI agents and automation, from one process to a full programme

I take on the whole job: finding the right process, building it properly and keeping it running. Pick the service that matches your problem, or start with a discovery audit if you’re not sure.

AI agents for operations

AI agents handle the parts of a process that need reading and judgement: invoices, emails, forms, applications and tickets. Wherever a mistake would be costly, I design them to propose rather than act, so a person approves the decision and the agent does the legwork.

Good fit when the work involves unstructured documents, varied requests or decisions that follow rules most of the time.

  • Copilot Studio
  • LLM APIs
  • RAG
  • Python
  • Power Automate
  • Airtable

Typical work

  • Document intake: invoices, applications and forms read, checked and keyed
  • Email and ticket triage, with drafted replies for a person to approve
  • Assistants that answer from your own policies and records
  • Copilot Studio agents that call your existing flows and desktop robots
  • Multi-step pipelines for research, enrichment and outreach, with approval gates

What you get

  • A working agent in your environment, not a demo
  • Approval steps and clear limits on what it can read and change
  • A log of every input, decision and action
  • Testing on your real examples before go-live
  • A runbook and handover for your team

Process automation (RPA)

Robots are still the most reliable answer for high-volume work with fixed rules, especially when the systems involved have no API. I build unattended and attended bots with proper exception handling, so failures land in a queue for a person instead of disappearing.

Good fit when the steps are fixed, the volume is high and the systems are old.

  • UiPath
  • Power Automate Desktop
  • Automation Anywhere
  • SQL

Typical work

  • Onboarding and KYC back-office steps
  • Card, account and policy maintenance at volume
  • Reconciliation between systems and reports
  • Bulk updates and data migration through legacy portals
  • Report generation and distribution

What you get

  • Process and solution design documents
  • Bots built to your platform’s standards, with retries and exception queues
  • An audit-ready log for every transaction
  • Test evidence and a hypercare period
  • Monitoring and support options

Power Platform apps and workflows

If you have Microsoft 365, you probably own most of what you need already. I build Power Apps and Power Automate solutions that replace spreadsheet trackers and approval-by-email with tools your team can rely on.

Good fit when your team runs on spreadsheets, shared inboxes and email chains.

  • Power Apps
  • Power Automate
  • Dataverse
  • SharePoint
  • Copilot Studio

Typical work

  • Approval workflows for purchasing, onboarding and change requests
  • Canvas and model-driven apps on Dataverse or SharePoint
  • Integrations between Microsoft 365 and your line-of-business systems
  • Error handling and status tracking that users can see
  • Governance basics: environments, solutions and data policies

What you get

  • Apps and flows packaged as solutions your IT team can support
  • Environment and connection set-up
  • Documentation and a walkthrough for admins and users

Automation discovery audit

A fixed-scope, fixed-price review for teams that know there’s manual work to remove but not where to start. It’s useful whether or not you build with me.

Good fit before any budget is committed, or when several teams are competing for the same automation effort.

  • Process mapping
  • ROI model
  • Roadmap

Typical work

  • Walk-throughs and interviews with the people doing the work
  • An inventory of candidate processes with volumes and time spent
  • The right approach for each: AI agent, RPA, Power Platform or leave it alone
  • Savings and effort estimates for each candidate

What you get

  • A prioritised roadmap with the first three candidates scoped
  • A business case you can take to finance
  • A recommendation on platform and licensing

Engagements

Ways to work together

Fixed prices wherever the scope allows. Most clients start with a discovery audit or a pilot and expand from there.

  • Discovery audit

    £X fixedTwo weeks

    • Walk-throughs with the people doing the work
    • Ranked list of automation candidates
    • Savings and effort estimate for each
    • Roadmap and business case for finance
    Book a discovery call
  • Agent pilot

    From £XFour to six weeks

    • One agentic workflow in your environment
    • Approval steps and audit logging
    • Tested on your own examples
    • Go or no-go review with real numbers
    Discuss a pilot
  • Build sprint

    From £XTwo to eight weeks per process

    • RPA or Power Platform delivery
    • Design, build, test and deploy
    • Two weeks of hypercare
    • Documentation and handover
    Discuss a build
  • Run and improve

    £X a monthNotice period, e.g. 30 days

    • Monitoring and fixes
    • Small changes every month
    • Quarterly review of savings
    • Priority response times
    Discuss support

Questions

Common questions

Do you work remotely?

Yes. I work remotely with clients in the UK and Europe, during UK and European business hours. Discovery sessions run over Teams or Zoom, and on-site visits can be arranged for larger projects.

How do you handle our data and security?

I build inside your environment, using service accounts with only the access each process needs. Credentials live in your vault or key store, never in code. For AI work we agree up front which data a model may see, and I use providers and settings that don’t train on your data.

Will an AI agent make decisions without us?

Only where you decide it should. By default, agents prepare, check and recommend, and a person approves anything with money, customers or compliance attached. As confidence builds, you can choose to let more run unattended.

Which platforms do you work with?

Microsoft Power Platform (Power Automate, Power Apps, Copilot Studio, Dataverse), UiPath, Automation Anywhere, Python and the major LLM APIs. I’ll recommend the platform that fits your licences and your team, and I’m happy to work with what you already own.

Who owns what you build?

You do. Code, flows, documentation and credentials are yours from day one and are delivered into your own accounts.

What happens if something breaks after go-live?

Every build includes exception handling and alerts, plus a hypercare period after launch. After that you can run it yourselves with the runbook, or keep me on a monthly support plan.

Not sure which service fits?

Describe the process on a 30-minute call and I’ll tell you which approach makes sense, including when automation isn’t worth it.