Automation should remove repetitive work, not add another complicated tool.
There's a lot of noise about automation at the moment, and most of it is aimed at selling you a subscription. The useful version is duller and more valuable: the small jobs your team repeats every week, done by the system instead. I'm Steve Bodman, and I've been building this kind of thing into business systems since 1997.
- Built into your system, not bolted on
- No extra subscription or login
- Visible when it runs, and when it fails
- Remote projects, UK-wide
Most of what a business loses to repetition needs no AI at all.
The time that disappears in a small business usually goes on things that follow an obvious rule: someone has to remember to send the reminder, re-type the details, produce the same document again, or check whether anything has been missed. None of that needs a model. It needs the system to know the rule.
AI does have a genuine place — mainly in drafting text a person then edits — and there's a section on that below. But if someone opens with AI before asking what your team actually repeats, they're selling rather than solving.
The same information typed into a second system because the first one can't pass it on.
Reminders that depend entirely on one person remembering them.
A document rebuilt from a template every time, from details already held elsewhere.
New enquiries sitting in a shared inbox until somebody claims them.
Renewals discovered after they've lapsed, because nothing was watching the date.
The same report assembled by hand, the same way, every single month.
Frequent, predictable, and driven by a rule.
That's the test. If a task happens often, happens the same way each time, and somebody could write the rule down on a single line, it's a candidate. If it needs judgement each time, it usually isn't — and automating it badly is worse than leaving it alone.
Enquiry routing
A new enquiry captured, categorised and sent to the right person with the context attached, instead of sitting in a shared inbox.
Reminders
Services due, quotes going cold, follow-ups owed and deadlines approaching — raised on the day they matter.
Scheduled emails
Confirmations, updates and summaries sent when a rule says so, with a record on the customer's file of what actually went out.
Renewal notifications
Contracts, certificates and subscriptions flagged in advance, so a renewal is a decision rather than a discovery.
Document creation
Quotes, invoices, certificates and job sheets produced from information the system already holds, rather than retyped into a template.
Recurring jobs
Annual services, monthly visits and scheduled inspections raised automatically on the right cycle.
Reporting
The figures you check regularly, produced and sent on schedule instead of rebuilt in a spreadsheet each time.
Status changes
When a quote is accepted, a job opens. When the job is signed off, it becomes invoiceable. The steps that always follow other steps.
Staff notifications
The right person told when something needs them, rather than everybody told about everything until they stop reading.
Follow-ups
The chase nobody enjoys: raised as a task, or sent automatically where that's appropriate, so it stops depending on memory.
Data moving between parts of a system
The enquiry becoming a quote becoming a job becoming an invoice, without anybody copying anything across.
Integrations with other systems
Where an external system offers an API, information can flow to and from it. Where it doesn't, it can't — that's checked early.
Where AI earns its place, and where it doesn't.
AI is genuinely good at producing a first draft of something a person then improves. It's not good at being trusted unsupervised with anything that reaches a customer, a price or a legal obligation. Those two sentences cover most of what needs saying.
Drafting text a person will edit
A first version of a description, a summary or an article — quicker to edit than to write from nothing, and reviewed before it goes anywhere.
Summarising something long
Turning a wall of notes or correspondence into something readable, with the original still there to check against.
Anything sent to a customer unread
A confident, plausible and wrong reply costs far more than the few minutes it saved. Drafts get reviewed.
Prices, quantities and rules
Anything that has to be exactly right belongs in ordinary code where it behaves the same way every time and can be tested.
This site runs on that principle. The article drafting on Millennium Software's own blog is AI-assisted: a draft is generated, then read and edited before anything is published. That's a real example of the boundary, and it's the same boundary I'd recommend to you.
Start with one task, not a strategy.
Automation projects go wrong when they start with an ambition instead of a task. Pick the single thing your team repeats most, automate that properly, and let it prove itself before adding the next one.
Name the task
Not "we want to automate admin" — the specific thing somebody does every Tuesday that they'd rather not.
Write down the rule
If the rule can't be written down, it isn't ready to automate. Working that out is often useful on its own.
Check what's possible
Where another system is involved, whether it allows access at all is established before anything is promised.
Build it where the information already is
Inside the system your team already uses, so there's no extra login, no extra subscription and nothing new to learn.
Make it visible
A record of when it ran and what it did, and a clear failure rather than a silent one. Invisible automation is how things get missed.
Then do the next one
Once it's saving real time and everyone trusts it, move on to the next task on the list.
Automation I've built and rely on.
These are Millennium Software's own systems, so the claims are things I can stand behind directly rather than outcomes reported by somebody else.
The system running this business
Renewals, recurring charges, support contract time and the working diary are handled by the system rather than by somebody keeping track, alongside automated email and content tools.
- Invoicing, renewals and reporting
- Support contract time tracking
- Automated email and content tools
- AI-assisted article drafting, reviewed before publishing
CoachFinder4U
A platform where the process runs itself: people are onboarded, verified, matched and billed, and told what's happening at each stage, without an administrator driving each step by hand.
- Onboarding and verification
- Matching between coaches and clients
- Subscriptions, credits and payments
- Automated notifications
Frequently asked questions
What kind of tasks are worth automating?
The ones that are frequent, predictable and follow a rule somebody could write down. Reminders, renewal notices, routing a new enquiry to the right person, producing a document from information already held, and moving data from one part of a system to another are all good candidates. Work that needs judgement each time usually isn't.
Is this the same as buying an automation tool?
No. A separate automation tool is another subscription, another login and another thing to maintain, sitting between systems that still don't really know about each other. Automation built into your own system has the information already, so there's nothing to connect and nothing extra for anyone to learn.
Do we need AI for this?
Usually not. Most of the time a business loses is lost to ordinary repetition, and ordinary rules fix it. AI is worth using for a narrow set of jobs, mainly drafting text, and even then a person should read it before it goes anywhere near a customer.
Where does AI genuinely help?
Drafting is the honest answer: a first version of a description, a summary or an article that a person then edits. Millennium Software uses exactly that on its own site, where article drafts are generated and then reviewed before anything is published. Nothing goes out unread.
Can you automate something between two systems we already use?
Sometimes, and it depends entirely on what those systems allow. If both provide an API or a usable export, moving information between them is often practical. If one doesn't, it may not be possible at all — which is worth establishing early rather than assuming.
What happens when something goes wrong?
Automation should be visible. Anything that runs on a schedule needs a record of whether it ran and what it did, and anything that fails needs to say so rather than fail quietly. That's part of building it properly, not an optional extra.
Will it put people out of work?
In small businesses that's rarely what happens. The work that gets automated is normally the part nobody wanted to do anyway, squeezed into gaps in the day, and removing it usually means the same people can handle more work rather than fewer people being needed.
Can automation be added to a system we already have?
Often, yes — and it's frequently cheaper than replacing anything. Whether it's practical depends on how the existing system was built and what access it allows, which is worth looking at before deciding either way.
Automating one thing, or rethinking the whole system?
Most projects cross over between these. If yours does, it's still one system — these pages are just different ways in.
Custom CRM Software
Customer records, enquiries, quotes and follow-ups shaped around your sales and service process rather than a generic pipeline.
Read moreBusiness Management Software
One system covering customers, jobs, scheduling, documents, invoicing and reporting instead of several that don't talk.
Read moreBooking Management Software
Availability, online booking, deposits, confirmations and reminders built around how your business actually takes bookings.
Read moreBespoke Software Development
How a custom project actually runs: scoping, staged development, your own live version to test, launch and ongoing support.
Read moreCase studies
The platforms Millennium Software has designed, built and still runs, including the system this business is run on.
See the workThinking it through first?
Tell me the task your team keeps doing manually.
Not a strategy — one task. The thing somebody does every week that they'd rather not, or the information that keeps getting typed in twice. Describe it in plain English and I'll tell you honestly whether it's worth automating and roughly what's involved.
You'll be talking to Steve Bodman, who will do the work — not a sales team, an account manager or an agency in the middle. If what you need is something off-the-shelf, I'll tell you that too.