01 · Documents
Read the paperwork
Invoices, delivery notes, bills of lading, statements. AI pulls out the fields and files them. Anything it isn’t sure about goes to a person.
Glasgow · AI, software and data
AI you can check.
We’re a Glasgow software team. We build AI into the systems you already run, so it reads the documents, sorts the inbox and answers questions from your own data. Its answers show where they came from. It runs privately where it needs to, a person checks what matters, and the price is fixed before we start.
Est. 2019 · Glasgow · You own the code, prompts and data
What AI can take on
Give AI the work your team shouldn’t be doing
Most teams use AI to reword emails or look things up. The real gains come when it’s built into how work gets done. Whether you need an AI strategy for the whole business or three or four jobs done properly, we start with the work that saves the most time and money. These are the jobs we build most often.
02 · Inbox
Sort the inbox and draft the replies
Enquiries are sorted, routed and answered in draft. Nothing is sent until someone on your team presses send.
03 · Answers
Ask your own data a question
Ask in plain English and get a number with a link to the screen it came from. It reads your systems; it can’t change them.
04 · Admin
Hand over the admin between systems
Agents move work from one system to the next: update prices, file orders, chase paperwork. Anything that spends money or talks to a customer waits for approval.
05 · Research
Keep watch on the public record
Monitor registers and filings for the companies you care about, and flag what changed, with the source attached.
06 · Reviews
Check the code and the agents
Already using AI to write code or run tasks? We review the code, the security and the agents themselves, and every finding is checked twice.
See it working
Real systems, not slideware
These are systems we’ve built and run. Client work is shown without names unless the client has said yes.
Paperwork captured at the loading dock
A phone photo or a copier scan becomes a searchable record, matched to the right customer. Scans it can’t split with confidence go to review.
AI that can read your CRM, and nothing more
Our own CRM, Tonic Desk, connects to Claude and other assistants through 19 read-only tools. There are no write tools, so the AI can’t change a record.
Where your data goes
You choose where the AI runs
Before we build anything, we draw your data’s route on one page and agree it with you.
Tier 1
A hosted model, through a UK or EU endpoint
The best quality for the least effort. Before any of your data goes to a provider, we check that its business terms rule out training on it, and we tell you which region does the processing.
Tier 2
Inside your own cloud account
The model runs in your AWS or Azure account, under your access controls and your logs.
Tier 3
A private model on hardware in the UK
Open-weight models on your hardware or ours. Nothing goes to an outside AI provider. Our own research and document reading run this way.
Storage location and processing location aren’t the same thing. We’ll tell you both, in writing.
How we keep it accurate
Accurate first, clever second
Tested on your examples
Before a pilot starts, we agree a set of your real examples and a pass mark. The AI has to meet it before it goes live.
A source behind every answer
Answers link to the record, page or screen they came from, so anyone can check.
A person approves what matters
Payments, messages to customers, changes to records: someone on your team signs them off.
Logged, budgeted, switchable off
We keep a record of what the system did, every workflow has a monthly budget with a hard stop, and there’s an off switch. For the most sensitive data we can switch off detailed logging, and we tell you where we have.
What else we build
AI is one part of the job. We build the rest too.
Most AI work sits on top of ordinary software: a database, an integration, a dashboard, a website. We build all of it, so the AI has something solid to stand on.
Software
Custom systems, integrations with the tools you use, and iPhone and Android apps.
Cloud and security
Hosting, least-privilege access and secrets kept out of the code.
Who we work with
How to start
Start small, with a fixed price
You don’t need to commit to a big project to find out whether AI helps.
-
AI opportunity audit
Two weeks, fixed price. We find the three jobs most worth handing over, and the ones that aren’t.
£1,500 -
Pilot
Up to six weeks on one job, with the test set and pass mark agreed up front.
From £6,000 -
Build
We make it part of your systems, with approvals, logs and budgets. Fixed price.
-
Run
Monitoring, model updates and a monthly cost report. You can leave at any time and take everything with you.
Planning AI across the whole business? Start with an AI strategy.
Prices exclude VAT where VAT applies.
We’ll tell you when AI isn’t the answer
Sometimes a £40 subscription does the job. Sometimes a simple rule beats a model. We’ll say so before we quote, because a system you can trust is worth more to us than a bigger invoice.
Who you’ll work with
Two people you can call
Steven Sharp
Co-founder and director, Dataface · GlasgowSteven spent over a decade building systems for UK financial institutions, including Standard Life, Scottish Life and Royal London, where a missed edge case shows up in a regulator’s report. He brings the same discipline to smaller businesses, without the enterprise theatre.
Marta Sharp
Co-founder and director of operations, Dataface · GlasgowMarta runs delivery, quality and client communication. She’s the reason you hear back quickly and see working software every fortnight.
Not sure where to start? Pick the problem.
Questions
Where does our data go?
You choose. A hosted model through a UK or EU endpoint, a model inside your own cloud account, or a private model on hardware in the UK. We write down where your data is stored and where it’s processed, and we put it in your contract.
Is our data used to train AI models?
Not by us. Before any of your data goes to a provider, we check that its business terms don’t let it train on your data, and we show you where that’s written. We don’t use your data to train anything unless you ask us to in writing.
Who owns what you build?
You do: the code, the prompts, the configuration and the data. We hand over the source and the infrastructure access at every milestone. If we set up one of our own products for you instead, such as Tonic Desk, you own your data and settings and can export them at any time.
What happens when the AI gets something wrong?
It will, sometimes, so we design for it. Uncertain cases go to a person. Answers carry their source so mistakes are easy to spot. And we track accuracy against a test set every time something changes.
What does it cost to run?
You get a fixed price for the build and an estimate of the running cost per 1,000 items at today’s model prices. After launch, you get a monthly report of what it actually cost, with a hard budget limit.
What happens when the model changes?
We pin model versions, re-run your test set before switching, and only switch when the results hold up.
Do we need AI at all?
Maybe not. The audit is there to find out. If a rule, a report or an off-the-shelf tool does the job, we’ll say so.
How long does it take?
An audit takes two weeks. A pilot takes up to six weeks. For a full build, we agree the timeline with your fixed price, and you see working software every two weeks.
Do you only do AI?
No. We build custom software, data systems and websites too. The AI usually sits on top of those.
Do you work outside Scotland?
Yes. We work with clients across the UK and in the USA, by video call and screen share, and in person in Glasgow when that helps.
Get in touch
Tell us the job you’d hand over first.
One paragraph is enough. We’ll reply with honest thoughts on whether AI can do it, and what it would cost.