AI & Automation
I start with the work, not the tool. Where a task is repetitive, well defined and safe to hand over, I automate it or add AI to it, with clear permissions, testing and a person in charge.
New v6 draft from the service brief, awaiting review
Where it usually starts
Repetitive admin
The same details are copied between inboxes, forms and spreadsheets every week. Often a well-built automation handles it with no AI involved at all.
The same questions, again and again
Customers ask what your site already answers. A chatbot can answer from content you have approved and pass everything else to a person.
Tools that do not talk to each other
Enquiries, bookings and records sit in separate systems. Where a multi-step task also needs judgement, an AI agent can prepare the work and a person approves it before anything is sent.
Automation is not always AI. Where a fixed rule does the job, I recommend the rule: it is simpler to run and easier to check.
When AI fits, and when it does not
Usually a good fit
- Defined inputs and a clear result someone can check
- Repeatable decisions that follow the same rules each time
- Source information that exists and is reliable, such as content you have approved
- A level of risk you can accept if a step goes wrong
- A named person who owns the task and reviews the output
Usually not
- A process nobody can describe yet, or that changes every time
- Data that is incomplete, out of date or unreliable
- Sensitive decisions with no person reviewing them
- Work that depends on professional judgement, such as legal, medical or financial advice
Before anything is built we agree four things: what data the tool may see and where it is processed, what it is permitted to do, how its output will be evaluated, and what happens when it fails.
How I approach it
There is no fixed package or timeline. Each step decides whether the next one is worth taking.
Understand the task
We map the task as it is done today: who does it, how often, and what a good result looks like.
Assess data, risk and alternatives
I look at the data it touches, what could go wrong, and whether conventional automation or a change of process would do the job without AI.
Define permissions and success
We agree the least access the tool needs, the points where a person approves, and the criteria that show it works.
Prototype and test on real examples
I build a small version and test it against your own examples, including awkward ones, before anything goes live.
Hand over, monitor and refine
The working version runs on accounts you own, with documentation, logs and a regular review, so a named person on your side can pause, change or switch it off.
Everything in AI & Automation
Each of these can be booked on its own.
- AI Automation & IntegrationI automate defined, repetitive tasks and connect AI tools to the systems you already use, with clear permissions, testing and a person in charge.
- AI ChatbotsI build website assistants that answer from your own approved content, say when they do not know, and hand over to a person.
- AI Agents with Human HandoverI design AI agents for narrow, defined tasks, with limited permissions, a person approving what matters and a clear way to stop them.
- AI-ready WebsitesI add practical AI to your site or workflow: assistants, search, content tools and automations that fit how you already work.
Search, websites and AI
How AI assistants describe your business is a search question, covered under SEO & AI Search Visibility. Adding practical AI features to your own website is AI-ready Websites. No file or markup guarantees how an AI system will describe you; the work is clear, accurate and well-structured content.
Questions about AI and automation
Draft answers for review
Will it use AI, or conventional automation?
Whichever suits the task. Many tasks are better handled by a fixed rule or an integration between tools, and I will say so. AI is added only where a step needs language or judgement that a person can check.
Can it connect to the tools we already use?
That depends on what each tool supports. I check the integrations and access each one offers before recommending an approach, and say plainly if a connection is not practical.
Who approves what it does?
You do. We agree in advance which actions need a person to approve them. Anything sensitive, such as sending to customers or changing records, waits for that approval.
What happens if it gets something wrong?
Every workflow has a point where a person takes over, an alert when a step fails and a manual way to finish the job. Testing on real examples is there to find mistakes early, and the regular review looks for what testing missed.
What data does it need, and where does it go?
Only the data the task requires. Before anything is built we agree what may be sent to a third-party AI service, where it is processed and what must never leave your systems.
Explore an AI or automation opportunity.
Practical AI and automation for defined tasks, with clear permissions, privacy and a person to hand over to.
Need continuing responsibility rather than a defined project? Ongoing partnerships