SPTech

Everyone says you should
be “using AI”.
Nobody says how.

We're a small team that helps ordinary businesses work out where AI genuinely saves time — and where it's a waste of money. No jargon, no jargon-shaped meetings.

Try it yourself, it takes about a minute

Question 1 of 3

The honest version

What it's actually good at.

Most disappointment comes from pointing it at the wrong job. Here's the split, in plain terms.

Six things that change

What actually changes in a business.

Not predictions — the six areas where we've seen the work genuinely shift, what each one looks like on an ordinary Tuesday, and what it asks of you first.

01

The conversations you keep having

Every business answers the same handful of questions hundreds of times a month — where's my order, do you deliver here, what's your returns policy. That volume is invisible because it's spread across everybody's inbox.

On an ordinary Tuesday

  • Enquiries arrive, get read and sorted, and land with the right person — with a suggested reply already attached for them to approve or rewrite.
  • The routine ones are answered from your own published information, with a link to where the answer came from so anyone can check it.
  • Anything unusual, angry or worth a lot of money gets flagged and pushed forward rather than sitting fifteenth in a queue.

First real results

4–6 weeks from starting

What you need first

A few months of real enquiries, and a decision about what a person must always see

The way this goes wrong: a system that replies fast and confidently to things it misread. The fix isn't better technology, it's deciding early which categories never get an automatic answer.

02

Paperwork stops being typed twice

Invoices, applications, delivery notes, claims, contracts. Somebody reads them and types what they say into something else. That job is slow, dull, and where most avoidable errors are born.

On an ordinary Tuesday

  • Details are pulled straight out of PDFs, scans and photographs into your spreadsheet or system — no retyping, no second person checking the typing.
  • Missing signatures, totals that don't add up and absent attachments are flagged before anyone opens the file.
  • Anything uncertain is held back for a person, with the doubtful field highlighted so checking takes seconds instead of a full re-read.

First real results

3–5 weeks from starting

What you need first

100–200 real documents, deliberately including the crumpled and badly scanned ones

Worth knowing: accuracy should be measured field by field, not document by document. “95% accurate” sounds fine until you learn the missing 5% is always the bank details.

03

What your company knows becomes findable

Most of what a business knows sits in old shared drives, finished email threads, and the heads of two people who've been there longest. New staff spend months learning who to ask.

On an ordinary Tuesday

  • Someone asks a question in normal words and gets an answer with a link to the document it came from, so they can judge whether to trust it.
  • New starters stop needing a colleague permanently on hand, which quietly gives that colleague their week back too.
  • Documents that contradict each other surface as a visible problem you can fix, rather than an argument two departments have every quarter.

First real results

4–8 weeks from starting

What you need first

A decision about which documents are the official ones — the hardest part, and it isn't technical

Be ready for this: pointing it at everything you own on day one mostly produces confident answers drawn from a policy that expired in 2019. Start narrow and expand.

04

You can see a bit further ahead

Stock, staffing, demand, cash. Most businesses plan these from last year plus a feeling. There's usually more signal in your own history than anyone has had time to extract.

On an ordinary Tuesday

  • Forecasts update themselves as orders come in, instead of being rebuilt by hand each month by whoever is best at spreadsheets.
  • Reorder and rota suggestions arrive with their reasoning shown, so an experienced manager can overrule them for reasons the data can't see.
  • When a pattern breaks, you hear about it in days rather than discovering it in the month-end numbers.

First real results

6–10 weeks from starting

What you need first

Two to three years of tidy history — genuinely the constraint here

The honest caveat: this is mostly statistics rather than anything exotic, and it needs real history. With less than about two years of data, don't start here — you'd be paying for a confident guess.

05

Mistakes get caught before customers see them

Most businesses find their errors the expensive way: a customer complains, or a number is wrong at month-end. The rules that would have caught it usually exist — in someone's head.

On an ordinary Tuesday

  • Every quote, order or claim is checked against your own rules as it's created, not audited weeks later.
  • The genuinely odd one — wrong by a factor of ten, a discount nobody authorised — gets held for a second pair of eyes.
  • Patterns in the errors become visible, which often points at a broken form or a training gap rather than a careless person.

First real results

4–6 weeks from starting

What you need first

Your rules written down — the exercise itself usually finds three you'd all forgotten

Expect friction early: it will flag things that are perfectly fine. Set the threshold too tight and people learn to click past the warnings, which is worse than having none.

06

The writing stops starting from nothing

Proposals, reports, product descriptions, job adverts, the same follow-up email for the four hundredth time. Nobody enjoys the blank page, and it's usually the task that gets pushed to the evening.

On an ordinary Tuesday

  • A first draft appears in seconds, built from your own past work, so it sounds like your business rather than like everyone's.
  • The job becomes editing, which is faster and far less draining than inventing — and better suited to the people who know the customer.
  • Work that used to be skipped because there was no time — the follow-up, the case study — starts actually happening.

First real results

2–3 weeks — usually the fastest win

What you need first

Twenty pieces you were happy with and five you weren't. The gap between them is the instruction.

One firm rule: a person reads it before a customer does. Every embarrassing AI story you've read in the news is a business that skipped this step to save thirty seconds.

How we help

Three steps. Then we're gone.

We're not trying to become a permanent line on your budget. The point is to leave you with something your own people can run.

We talk

A free half-hour call. You describe what eats your week. We tell you honestly whether this is something AI helps with — and quite often the answer is no, and that's the end of it.

Costs nothing

We try one thing

We pick the single most annoying task and build something small that does it, tested on your real work. You see it working — or not — in weeks, not quarters.

Usually 4–8 weeks, agreed price

We hand it over

We write down how it works in language your team can read, show them how to run it, and stay on for a few weeks while they drive. Then we go.

The finish line, in writing

Money and time

What it costs, roughly.

Most consultancies won't put numbers on a website. Here's the shape of it, so you can decide whether to bother calling. Every project gets a fixed price agreed in writing before it starts.

StageWhat happensHow longWhat you pay
First conversation

Half an hour on a call. You describe the problem, we say whether there's anything here — including when the answer is a £30/month tool and no consultant.

30 minutes
Free
Look properly

A short piece of work where we sit with the people doing the job, look at the real data, and write you a plan: what to build, what it'll take, what to leave alone.

1–2 weeks
5000 - 100000
Build one thing

One task, taken all the way into daily use — including the checking, the fallbacks, and the training so people actually use it.

4–8 weeks
5000 - 100000
Hand it over

Written instructions in plain language, named owners, and a few weeks of your team running it while we watch. Then we stop billing you.

2–4 weeks
5000 - 100000
Running costs

What it costs you per month afterwards, paid to the software providers directly — not to us. We size this before you commit, never after.

Ongoing
Yours, not ours

Makes it cheaper and faster

  • The work already lives in a spreadsheet or a real system
  • One person can approve a change to how the job is done
  • You have plenty of past examples, including the bad ones
  • A wrong answer is annoying rather than dangerous
  • You've picked one task, not a category of tasks

Makes it slower and dearer

  • Nothing is written down and the knowledge is in two people's heads
  • Data can't leave your building for legal or contractual reasons
  • Several departments have to agree before anything changes
  • Mistakes carry real consequences — money, health, someone's job
  • The old system is ancient and nothing can talk to it

Fair questions

Things people ask us.

Often, yes — and we'll say so. If you're a handful of people, the honest answer is usually that a tool you can subscribe to for a small monthly fee will do the job, and you need an afternoon of setup rather than a consultant. We'll point you at the tool on the free call and send you on your way. It costs us nothing and it's how we'd want to be treated.

The work we're good at removing is the work nobody wanted: retyping things, copying between systems, hunting for a document, writing the same email for the four hundredth time. If your plan is to cut headcount, we're probably the wrong people — that tends to go badly, because the staff you're cutting are the ones who notice when the system is wrong. What tends to work is the same people getting through more, with less of the dull part.

This is the right thing to worry about and it shapes the whole design. There are options: keeping everything inside your own systems, using providers who contractually don't train on your data, or stripping identifying details before anything is sent anywhere. Which one you need depends on your industry and your rules. We work this out in the first week rather than discovering it late, and we put it in writing.

It will, sometimes. That's not a reason to avoid it — it's a reason to design for it. Anything we build has a clear answer to three questions: how does a person spot a mistake, how do they override it, and what happens to the ones it isn't sure about. If a task can't survive an occasional wrong answer, we either put a person in the middle or we don't automate it at all.

We agree the measure before we build, not after — usually something you already count, like how long a job takes or how many get sent back. Then we take a reading of how things stand today, so there's something to compare against. If the number doesn't move, that's a real result and we'll tell you rather than pointing at how clever the technology is.

No — and it's the reason we write things the way we do. You don't need to understand how any of it works internally, any more than you need to understand a payment terminal. You do need to be able to tell us what good work looks like in your business, because that's the part no outsider can supply. That's the expertise we need from you, and you already have it.

No pressure

Have a chat with us.

Half an hour, free, no slides. Tell us what your week looks like and we'll tell you straight whether there's anything here for you.

Email us Or call: 9063515511

We answer within a working day. If we're not the right fit, we'll tell you who is.