Use cases
Ailligence use cases: typical problems, industry by industry, and how custom software, integrations and AI solve them. For anonymised real cases from my clients, see the AI Code Check-up.
-
An AI assistant answers the users of a web application using only the client's documents. I rebuilt the search with Anthropic's Contextual Retrieval: hybrid search by keyword and by meaning, with context inside every passage. Search in 20-30 milliseconds across about 30,000 passages, about one cent per answer.
-
I moved the history into the new software with imports that can be re-run, and connected it to the company's ERP. Today the two stay in step without any manual work.
-
I built a system that automatically sorts payslips uploaded in bulk into each employee's documents and records proof of receipt: a payslip can only be downloaded after it has been opened. It also works as an app that installs on the phone.
-
I moved more than seven thousand accounts, with their permissions, from one platform to another, with automatic email sending blocked during the import.
-
Ahead of a launch with more than 1,500 people expected at once, I fixed a bottleneck in the code that was slowing down sign-ups and moved the platform from a single server to an infrastructure that scales on its own. Sign-ups 86% faster, no errors in the load tests, launch on the planned date.
-
I took over a web application written by others and used every day by several companies. It was fragile and had no tests. Instead of rewriting it, I made the way it is changed reliable: re-runnable scripts, tests on a copy of production, a guide for every release. From there the application grows one piece at a time.
-
My client wanted the customer area on the web and also in the Apple and Google stores. I published the web app, originally built on Lovable, as an iOS and Android app with Capacitor, and in front of the company system I put a Laravel layer that standardises the responses and holds up when the system doesn't respond.
-
The management reporting took staff costs from an external file, and one day a column became two. I adapted the import, added employee matching with manual confirmation and a rate history: margin and average revenue per work centre are now calculated over the right period.
-
Exit alerts weren't arriving and the alerts about expiring documents were wrong. I found the cause by analysing the data received by the server, then rebuilt the flow in the iPhone and Android apps.
-
I built an AI assistant that answers by voice and remembers over time, with security designed before the features: sensitive data masked, credentials kept away from the AI, explicit consent for every change.
Facing a similar problem?
Tell me what you need: I reply myself, usually within one working day.