• 20 Nov 2026, 5:05 p.m.

    That sounds a bit like herding cats to me.

  • 20 Nov 2026, 5:09 p.m.

    I'm taking the view of being proactive in AI - so being one of the later ones to go - my hope is that we have a lot of technophobe clients who may keep me going on a declining level for a few years.

    The issue we are struggling with is making sure we have the correct GDPR policies in place - that is a job for Feb - probabyl getting a consultant - or a local AI model

  • 20 Nov 2026, 5:11 p.m.

    I understand the Department Of Education are changing the nursery rhyme to
    Old MacDonald had a data farm
    AI AI
    Oh...

  • 20 Nov 2026, 7:24 p.m.

    Isn't the main thing where the data us stored? A friend in a similar field to yours has been down this road. Had to get an enterprise license which was a wedge, upfront.

  • 20 Nov 2026, 7:27 p.m.

    You need to make sure some cunt isnt letting an unauthorized llm loose on your data. So easy to add a connector and then....

  • 20 Nov 2026, 7:36 p.m.

    In all seriousness, double down on learning AI prompt engineering and operational deployment. Most fuckers are faking it and theres a ton of money to be made

  • 20 Nov 2026, 8:29 p.m.

    We are looking at running a sandboxed local model for sensitive data.

    How much did the enterprise option cost?

    Copilot is supposed to be UK storage and processing this year

  • 20 Nov 2026, 8:31 p.m.

    I did think if we developed a local service with the right set up, OCR, data validation, analysis, it may be a marketable product

  • 20 Nov 2026, 8:37 p.m.

    We run our own models so that no client data ever leaves our "premises" and no external models are being trained on it. You then also get infinitely cheaper costs I think, as you aren't buying credits from the vendors all the time. Least that is my understanding of it.

  • 20 Nov 2026, 8:47 p.m.

    What models do you run? I imagine quite big ones as you will have a far bigger budget