Independent systems engineer
I’m David Bentley. I look at the systems a business already runs and build the piece that is missing.
You speak to me. I do the work.
The useful job is not always a tool. It might be course information that is easier to navigate, or an enquiry form that asks the right questions first time.
Course information
For a specialist aviation training provider, I built a Course Finder that narrows the existing course range by role and goal. I also delivered a Training ROI Calculator.
For the same provider, the knowledge hub redesign was in client review as of 16 September 2026.
No measured adoption, booking gains, time savings or realised return is claimed.
Shopify and Klaviyo records
For a UK wellness brand, I traced an account-wide suppression event dated 4 April 2025. It had gone unnoticed for fifteen months when I investigated it in July 2026.
I recovered and verified 2,151 profiles within the platform.
What that establishesPlatform status, not permission to contact anyone, delivered messages or sales. Sales impact was not measured.
This is my case account, not an independent audit. The underlying private records are not reproduced. It does not establish that another business has the same problem. The fifteen months describes that account before the investigation, not any individual person’s messaging history.
About AI
AI helps us research and draft. People make the decisions and check the result. There are three situations, and the answer is different in each.
01 · The demonstration finder
The rules-based finders prepared for this letter campaign work from a table of routes in the visitor’s browser. Suggested course routes refer to captured information from your pages. Where that information does not establish an answer, the finder asks your team to confirm.
The enquiry is assembled from the visitor’s answers. The same answers, with the same table, give the same result.
What it sees: the visitor’s answers, in their own browser. Nothing is sent to an AI service.
02 · Building from public pages
We use AI tools to read published course information and draft the route table. David reviews the build. On new client builds, Orietta checks the agreed facts against their sources and follows the customer journeys in scope.
What it sees: published course information and review instructions.
03 · A job using private data
The project’s AI model runs on a computer you own or in an account you control. Your private business data is not sent to a model provider’s AI service. You control the access, and you can switch it off at any time.
How: we use open-weight models. These are models whose files can be downloaded and run privately.
The answers become the agreed checks for the build and the plan for keeping it useful.
Checks find mistakes. They do not guarantee that every answer is correct.
Running AI on your side settles who holds the data. It does not make an answer right.
WhatsApp and email are separate services. They sit outside any private setup.
Control and handoverTell me about one step. A sentence or two is enough. Please leave out passwords, customer records, learner data and account access.
First exchange free. No call required. I will not chase a reply.
Before a build, I’ll set out the scope, what I’d need from your team, and the fixed price. We agree the payment terms and availability before work starts.
See how to start