Disclaimer

There are no photographs on this site

Every image here is a drawing, generated from code we wrote: the desk scene, the diagrams, the brand mark and the illustrations inside each section. None of them depicts a real person, a real office or a real client project, and no client premises or client work appears anywhere on this site.

We mention it because a picture of somebody at a desk implies a photograph of the person you would be hiring, and that would be a claim we cannot make. Where you see a figure at a screen, it is a line drawing standing in for the work described, not a record of it.

What AI systems can and cannot be relied on to do

This matters more than any other line on this page. Large language models produce likely text, not verified truth. A well-built retrieval system, a tight prompt, a strong evaluation suite and human review reduce the error rate a long way — they do not take it to zero, and anyone telling you otherwise is selling something.

Any system we build is measured against an accuracy bar agreed with you in advance, and it is designed so that the failure cases are visible and recoverable rather than silent. Deploying it into a workflow where a wrong answer causes harm without a human in the loop is a decision only you can make, and the responsibility for that decision stays with you.

Model providers also change their models. Behaviour that passed evaluation in one month can shift in the next. That is why evaluation suites are part of the build rather than an optional extra, and why ongoing monitoring is worth paying for.

No guarantee of results

Timeframes, cost ranges and improvement figures on this site describe what has been achievable, not what is promised to you. Outcomes depend on your data, your systems, your team's availability and factors outside anyone's control.

Where a case study or figure appears anywhere on this site, treat it as illustrative of an approach rather than as a prediction of your result.

Accuracy and availability

Content is reviewed periodically, but the field moves quickly and pages can fall out of date between reviews — particularly anything naming a specific model, price or provider capability. No warranty is given that the site is complete, current or error-free, and it may be unavailable at times without notice.

If you spot something wrong, please write; corrections are made quickly and gratefully.

External links

Links to other sites are provided because they were useful at the time of writing. We do not control them, do not monitor them, and am not responsible for their content, accuracy, security or privacy practices. A link is not an endorsement, and following one takes you outside this site's privacy policy.

Testimonials and examples

Any quotation, client comment or worked example is presented as given or as encountered. Individual results vary with circumstances, and nothing in an example should be read as a typical or expected result for your project.

Your responsibility for compliance

Whether a system you deploy is lawful in your sector and jurisdiction is your responsibility. That includes data protection, sector regulation, the EU AI Act's obligations where they apply to you, accessibility duties, and any duty to tell your own users that they are interacting with an AI system.

We will raise constraints we are aware of and design around them where we can. That is engineering input, not a compliance sign-off, and it does not transfer the obligation to us.

Limitation

To the fullest extent the law allows, no liability is accepted for loss arising from reliance on anything published on this site. Liability for paid work is governed by the terms of service and your signed project agreement, which set the caps that actually apply.