Services

AI consultancy services for businesses that have to get it right

Minibit, Steven Minnikin’s AI-first IT consultancy, offers four services that take a business from “we should be using AI” to AI running safely inside the systems it already has: AI-readiness discovery, safe AI rollout, integration and agent build, and data, MI and migrations. Ongoing support keeps it working.

01 · AI-readiness discovery

Know what AI can do for you before you spend on it

A short, fixed-fee look at how work actually flows through your firm, where the hours go, and where the risks sit. I watch real cases rather than a demo, talk to the people doing the work, and look at the data behind it.

You get a written report with a prioritised list of opportunities, the controls each one needs, costed options, and a recommended first pilot. You can act on it with me or without me.

Fixed fee2 to 4 days over a fortnightWritten reportCosted optionsNo obligation
Who it suits
  • Directors who are being asked “what is our AI plan?” and want a real answer.
  • Compliance leads who suspect staff are already using consumer AI tools.
  • Operations leads who know where the time goes but cannot prove it.
  • Firms about to change systems, or that have just been acquired.
02 · Safe AI rollout

AI your staff can use without anyone wondering where the data went

Most firms already have shadow AI: staff pasting client details into whichever chatbot they like. A safe rollout replaces that with approved business tools under a data-processing agreement, a short written AI policy, and assistants set up for your own templates and processes.

You get the tools configured, the policy written and agreed with your compliance function, a sign-off step for anything that reaches a client, and hands-on training for staff. For regulated firms the policy covers Consumer Duty, record keeping and senior manager accountability.

AI policyApproved toolsDPA in placeStaff trainingHuman sign-off
Typical first assistants
Document first draftsReports, letters and summaries from your templates and file notes
Reviewed
Meeting and call notesStructured notes and actions, filed to the right record
Reviewed
Inbox triageSorted, summarised and drafted replies waiting for approval
Reviewed
Checklists and QAFiles checked against your own criteria before they go out
Evidenced
03 · Integration and agent build

Agents, connectors and assistants inside the systems you already run

When the off-the-shelf tools run out, I build. MCP servers and Claude connectors secured with OAuth 2.1 so authorised staff can work in your systems from an AI assistant. Automations between the systems that do not talk to each other. Voice agents that answer the phone, book and hand off. Agents in your delivery pipelines.

You get working software with the controls built in: permission checks, approval steps, audit logs, spend caps, prompt-injection defence and evals that prove it still works after every change. Code you own, on standard infrastructure.

MCP serversOAuth 2.1 connectorsVoice agentsLangGraph and RAGEvals and guardrailsAudit trail
Built and shipped
Mini StepsGoal planning with owner-approved actions and end-to-end voice interaction through the MCP connector
Private beta
Ship KeysLaunch-readiness checks for AI-built apps, as a Claude connector and Claude Code plugin
Private beta
Mini CallsPhone and web voice agent with secured actions and multi-agent hand-off
Built
Mini OpsWeekly AI competitor reports for small businesses
Prototype
04 · Data, MI and migrations

Clean data, reports that stay current, migrations that reconcile

AI is only as good as the data under it, and most books are messier than anyone admits: client or product linking issues, duplicates, gaps in the MI, and reports rebuilt by hand every month. I fix the data, build reporting that lives inside the system, and run migrations that are checked record by record.

You get every record linked to the right client, duplicates and gaps resolved, management information you can trust, and, when you change systems or acquire a firm, a migration with a reconciliation report that proves the numbers match.

Intelliflo OfficeReconciliationMI and reportingMigrationAcquisition integration
Done before
  • Led the Lowes post-sale integration into the acquiring group: data migrated, platforms merged, legacy systems switched off.
  • Client book reconciliation and live reporting inside Intelliflo Office.
  • A 44-application, 45-database estate run for 80+ staff and 5,000+ clients.
The process

How an engagement runs

Four steps, each one small enough to stop after. The aim is a measured pilot within six to eight weeks of the first call.

1

Talk

30 minutes

A call about your priorities and what is slowing the team down. You leave with a straight answer on whether AI is the right lever, and if so where.

2

Discover

2 to 4 days

I map how the work actually flows, where the time goes and where the risks sit, then write it up with costed options.

3

Pilot

4 to 6 weeks

A small rollout on live cases with a person signing off every output. Hours saved and issues caught are measured before anything grows.

4

Run

Monthly

Monthly tuning as your process changes, a short report on what AI did and what it saved, and someone to call when something breaks. Stop on a month’s notice.

Ongoing support

Someone to call, and a monthly note on what the AI actually did

Processes drift, models change and staff find new uses. Support keeps the assistants tuned, the policy current and the evidence flowing, with a short monthly report on time saved and issues caught. It is priced after the pilot, sized to what is running, and you can stop it on a month's notice.

How I price
  • Discovery is a fixed fee agreed before I start.
  • Pilots and builds are quoted from the discovery report, with the options costed side by side.
  • Support is a monthly fee you can stop on a month’s notice.
  • No lock-in. Standard tools, open protocols and code you own.
FAQ

Questions about the services

How long does AI-readiness discovery take?

Two to four days of my time, usually spread over a fortnight so I can watch real cases rather than a demo. I am Steven Minnikin, and I do the discovery myself rather than sending a junior. You get a written report with costed options at the end.

Can we act on the discovery report without you?

Yes. The report names the tools, the order of work and the costs, so your own team or another supplier can run with it. There is no obligation to continue.

Do you build, or only advise?

Both. I design the approach and build it: assistants, automations, MCP servers, connectors, voice agents and data pipelines. My independent products demonstrate agent orchestration, voice interaction, approval workflows and cost controls.

What does ongoing support cost and can we stop it?

Support is a monthly fee agreed after the pilot, sized to how much is running. You can stop it on a month's notice. Everything is built on standard tools and code you own, so stopping does not strand you.

Do you work remotely?

Mostly, yes. I am in Newcastle upon Tyne and work with firms across the UK over video, with on-site days for kick-off, training and anything that is easier in a room.

Which AI models do you use?

The one that fits the job and your data rules. Most of my builds use Claude for reasoning and drafting, with other models where they are stronger or cheaper. Prompts and logic are kept model-agnostic so you are not locked to a provider.

Not sure which service you need?

That is what the first call is for. Thirty minutes, no cost, and a straight answer on where to start.

steve@minibit.co.uk · Newcastle upon Tyne, working across the UK