Why open source matters when you are small

You own it

Open-weight models and self-hosted tools run on a machine you control, in your office or on a UK server. No per-seat fees, no price rise when you add staff, nothing that disappears when a vendor changes its mind.

Your data stays yours

Customer lists, quotes and case files never leave a system you control. For a firm with confidentiality duties, that is the difference between being able to use AI and not.

It can be checked

Open code can be read, audited and fixed. STG can review it. A closed tool has to be taken on trust, and the terms can change after you have built your week around it.

What we actually do

  1. AI audit and roadmap

    An honest look at where AI would actually save you time, and where it would not. You get a short written plan, in order of payback, with the open-source option and the paid option side by side where both exist.

  2. Tool implementation

    Set up properly, with the rules written down first: what data goes where, who can use what, and what to do when it gets something wrong. Reviewed alongside STG before it goes live.

  3. Team capability uplift

    Training for the people who will actually use it, not just the owner. Across the UK, 32% of workers have had any AI training at all (ONS and CIPD data, 2026). The firms that close that gap inside their own walls are the ones that get the benefit.

The training figure comes from the Numbers page of our journal, where every statistic we use is listed with its source.

What it looks like in practice

  • A private assistant for your own documents

    An open-weight model running on your own server, answering questions from your quotes, manuals and past jobs. Nothing is sent to a third party.

  • Automation you can see

    A workflow tool you host yourself that moves an enquiry into your CRM, drafts the reply and books the job, with every step visible and editable by you.

  • Software you own

    A quoting, booking or job-management system built for how you work, on open foundations, with no per-seat fee and no lock-in.

  • AI in the tools you already have

    Sometimes the right answer is the AI feature already inside your accounts package or email. When that is true, we say so and save you the project.

When open source is the wrong call

Open source is not free. Somebody has to run it, update it and back it up. If you have nobody to do that and no budget for us to, a well-chosen paid service can be the better decision. We will tell you which, and why, before you spend anything. What we will not do is rent you something you could have owned.

What it costs

Fixed fees, published before we start. The advisory work is priced on the advisory page of our journal, Terrified of AI. Bigger builds are quoted in plain numbers before any work begins. There is also public money for this: the Scottish Government funds free AI training and support for SMEs until March 2027, and we wrote a plain-English guide to getting it.

Questions we get asked

Is open-source AI good enough for a small business?

For most day-to-day jobs, yes: drafting, summarising, answering questions from your own documents, sorting enquiries. Where a task genuinely needs the largest frontier model, we say so and set that up safely with the rules written down, rather than pretend the open option is always best.

Do we need our own server?

Not always. A capable office machine runs smaller models well for one or two people. A rented UK server suits a team. We size it to the job and tell you the running cost before you commit to anything.

Who keeps it running?

Either your own team once they are trained, or us on a fixed monthly fee. What we will not do is hand over something nobody can maintain. If neither option fits, we will tell you a paid service is the better decision.

What about security?

Every setup is reviewed alongside Swanswater Technology Group (STG), the specialist cyber security company we work closely with. The rules on data are agreed and written down before any tool goes live, and self-hosting means there is less to trust and more to check.

Start with the honest look

Five minutes on the free diagnostic tells you whether AI is the thing to fix first, or whether something more basic is holding you back.