AI solutions for small and medium businesses
There is a job in your business that eats a day a week and lives in a spreadsheet. We build the software that does the reading, the sorting and the chasing — properly, for a company your size, and honest about what it cannot do.
Scroll
Why this is still a problem
AI was built for companies with a data team. Most businesses have someone who is good with spreadsheets.
Enterprise platforms assume a procurement team, an integration budget and a twelve-month rollout. A thirty-person business has none of those. So it gets the cut-down edition of a product designed for a company with a thousand staff, or a chatbot bolted onto a website and called a strategy.
Language models turned document comprehension into a commodity almost overnight. That on its own is not a product. What matters is what happens when the answer is uncertain: scoring it, showing where it came from, and putting a person in front of it when the stakes justify one. That layer is the work, and we have already built it.
We are not a consultancy billing you to discover this from scratch. Ingestion, extraction and confidence routing already exist and are already tested. You pay for the part that is actually yours, which is why a build takes weeks rather than quarters.
What we build
If your version of the problem is not here, it is probably a variation on one of them. The first conversation is about working out which.
Contracts, invoices, purchase orders, statements of work, scanned paper. Read once, turned into structured data you can search, filter and report on — with the source line quoted next to every value so anyone can check it.
The report that takes someone two days and is out of date by the time it lands. Same numbers, same format, assembled automatically, with the workings visible so nobody has to take it on trust.
Built from your own previous work and your own pricing rules, not invented from nothing. A first draft in minutes that a person edits, instead of a blank page at five o'clock on a Friday.
Every supplier, every renewal, every obligation, in one place. Dates plotted by the deadline to give notice rather than the date the contract ends — because the first one is the one you can still do something about.
Inbound email and form submissions read, classified, prioritised and routed, with the ones that need a person flagged as needing a person. No auto-replies pretending to be human.
The spreadsheet with four spellings of the same supplier. The export that never quite matches the system it came from. Unglamorous, and usually the thing standing between you and everything above.
A worked example
An illustration, not a customer. We have no case studies yet and will not invent one — but this is the shape of the work, and the shape of the conversation.
What we would not claim. That it read all 140 perfectly. Some scans are bad, some contracts are strange, and a handful always need a human. The point is not that the machine is never wrong. It is that when it is unsure, it says so, and shows you the line it was looking at.
The engine
The hard part was never getting a model to read a document. It is everything wrapped around the moment it gets something wrong — and that is the part we have already paid for.
PDFs, Word, Excel, CSV, meeting transcripts and photographs of paper. Scanned documents are read on the device with optical character recognition, because in practice a good share of what a small business holds is a photograph of a contract rather than a file.
No figure appears without a confidence score, the reasoning behind it and the quoted line it came from. If the system is unsure, that value is not quietly written into your records as though it were fact.
High confidence passes automatically. The middle band gets a quick check. Low confidence goes to full review and never reaches a report unverified. For a business without a team to catch mistakes, this is the layer that decides whether the whole thing is usable.
The screens, registers, searches and exports your people will actually open. This is the layer that is yours; the three beneath it are inherited, tested and shared. It is the whole reason a build takes weeks instead of quarters.
Honestly compared
| Option | When it is the right call | What it costs you |
|---|---|---|
| Do nothing | The job is annoying but small, and nobody is making decisions on bad information | Nothing, until the day a renewal slips. Genuinely the right answer more often than a vendor will admit |
| An off-the-shelf platform | Your process is standard and you can bend to fit theirs | Per-seat pricing that scales badly for a small team, and a rollout designed for a company ten times your size |
| Hire someone | The work needs judgement more than it needs speed, and it will keep growing | A salary, recruitment, and the risk that the knowledge leaves when they do |
| Build it in-house | You already have a developer with time and the problem is genuinely unique to you | Their time, and the part everyone underestimates: what to do when the model is wrong |
| Us | The job is repetitive, document-heavy or report-heavy, and the cost of a quiet mistake is real | A fixed build fee and a monthly run cost. You own the result |
Our own product
ContractIQ is the proof that the engine works, and the thing we point at when someone asks what this looks like finished.
Contract intelligence for teams that buy things. It reads contracts, purchase orders, statements of work, tracking spreadsheets and meeting transcripts, then tells you what you are actually committed to — before you walk into a renewal conversation without the facts.
Screens from the ContractIQ Platform. Figures shown are illustrative example data, not a customer’s portfolio.
Working with us
Ingestion, extraction and confidence routing are written once and shared across everything we do. You pay for the part specific to your business, not for us to rediscover the rest.
A model's confidence is not measured accuracy, and we will not quote you a percentage until there is a properly labelled test set behind it. Plenty of vendors will. Ask them for the workings.
We would rather do one thing that gives you back a real afternoon every week than a platform that does nine things adequately. The second one stops being opened by month three, and everybody knows it.
When something cannot be reached, is not yet verified, or the model was cut off mid-answer, the screen says so in a sentence your team can act on. Silent degradation is how trust in this category dies.
Whatever we build for you is yours. The code, the data and the outputs. If you stop working with us, you keep it, and we will hand over everything needed to run it elsewhere.
About CodeIQ Holdings
Most artificial intelligence platforms are priced, scoped and sold to businesses with dedicated data teams, large integration budgets and twelve-month rollout plans. If you are a small or medium-sized business you usually get left with two bad options: the watered-down version of an enterprise tool, or a generic chatbot bolted onto your website.
We don't sell off-the-shelf software and leave you to work out how it applies to your business. We find solutions to your specific operational bottlenecks.
We started CodeIQ Holdings with a single guiding principle: AI should eliminate the tedious operational work that still lives in spreadsheets, scattered inboxes, multiple OneDrives and one person's head.
After years of leading digital transformation and operational automation at enterprise level, our founder, Raz, saw that SMEs were being left out of the AI conversation entirely. The technology to automate document comprehension, quote drafting and reporting already existed — it simply was not being packaged in a way that made sense for a smaller team.
We bridge that gap. We focus on the pain points that eat up your Thursdays. Whether it is messy data clean-up, supplier visibility, or reporting that finally updates itself, we look at the problem first and point the technology at it second.
We work on a hybrid model that saves you time and money. We are not a consultancy billing you by the hour to build a system from scratch, and we are not a software company forcing you to change how you work to fit our product.
Instead we use a shared, pre-built engine that does the heavy lifting — ingestion, extraction and judgement, set out in full further up this page. Those three layers are already built and already tested.
So the only part built from scratch is the last one: your surface. The screens, registers and exports, designed around how your people actually work. That is what makes a build take weeks instead of quarters, and it is why you are not paying us to rediscover the rest.
We are deliberately honest about what AI can and cannot do. When our system is uncertain it does not guess — it shows its workings and asks a human.
We do not hide behind unmeasured accuracy percentages, and we do not train public models on your private data. Both of those are commitments in our contract, not slogans on a website.
Short commitments. Working software early. Real problems solved.
Before you ask
Usually the opposite. The smaller the team, the more of the operational load sits with one or two people, and the more it costs when they are unavailable. The question is not headcount, it is whether the same task is done the same way often enough to be worth building for. If it happens once a quarter, it probably is not.
They are encrypted in transit and at rest, held in your own workspace, and never used to train any model — ours or anyone else's. Our AI providers are engaged on terms that prohibit training on customer content. If you leave, you export everything and we delete it. The full position is in our data processing agreement, which you can read before you speak to us.
It will, sometimes. Every value carries a confidence score and the quoted line it came from, so a person can check it in seconds rather than re-reading the document. Anything the system is unsure about is routed to a human and never written in as fact. We would rather show you an uncertain answer honestly than a confident one that is wrong.
No. Most of what we build sits alongside what you already use and reads what you already have — files, exports, spreadsheets, email. Where a direct integration genuinely helps we will say so, but "rip out your systems first" is not a project we take on.
You own the software we build for you, your data and the outputs. The shared engine underneath — ingestion, extraction, confidence routing — stays ours, because it is the same engine every customer benefits from. If you stop working with us, you keep your part and we hand over what is needed to run it.
The build is fixed price and quoted before we start, so the cost cannot run away from you. It is paid half on starting and half on acceptance, which means we carry the risk of finishing. The monthly run cost cancels on 30 days with no minimum term. If we get partway in and conclude it is not going to be worth it, we will tell you rather than keep going.
A fair question, and the honest answer is that early customers get disproportionate attention, direct influence over what gets built, and a price that reflects the risk they are taking. If you need a vendor with a hundred references and a SOC 2 report, we are not that yet, and we will tell you so rather than let you find out in a due diligence questionnaire.
The United Kingdom. We work with businesses across the UK and most of the work happens remotely, with calls and screen shares. If you would rather meet in person, that is usually possible — tell us roughly where you are and we will say honestly whether it is practical.
Get in touch
One conversation, no charge, and an honest answer about whether it is worth automating at all. Sometimes it is not, and we will say so.
We reply to everything, usually within one working day. We do not add you to a mailing list, and we do not pass your details to anyone. What we do with them is set out in our privacy policy.
We use cookies that are strictly necessary to make this site work. We would also like to set optional ones to understand which pages are used. You can accept, reject, or choose. Full detail in our Cookie Policy.