AI is part of everything Elephas does: the work we deliver for clients, and the work of running the company. That is where our advice comes from — daily practice, not a slide deck.
Where we use it
Client delivery
Every project: AI agents build to a written specification, and an experienced engineer reviews every change before it ships. Two legacy systems recently went from plan to live in under a week each.
Read how we work →Sales, marketing and finance
Proposals, marketing material and our own finance process and dashboard all run with AI. It's how a small company keeps the administration to a minimum.
Design to code
A process that turns Figma designs into pixel-perfect websites, with every view checked against the design by automated screenshot comparison.
Read more →Websites and content
Umbraco websites built with AI from start to finish, including the content management.
Keeping systems current
Across the systems we support, AI does the heavy lifting of framework upgrades and fixing vulnerable package dependencies — work that used to be postponed because nobody had time for it.
Data
Cleansing and reconciling data at scale, dry run first. One recent pass made 1,321 corrections across 24 buildings with no errors.
Read the case study →Learning
We think AI will change how people learn as much as how they work. To test that, we built a Swedish practice app for advanced learners over a weekend, and it is now in daily use.
Learners write their answers in their own words. Claude grades each answer, explains what was wrong and records the type of mistake, and a spaced-repetition schedule brings weak points back at the right time. We chose the grading model by testing several against answers with known verdicts, and the app logs what every graded answer costs. That is the same discipline we bring to client work: measure first, then decide.
What we've learned
The right technology, not the familiar one
When AI writes most of the code, the cost of working in an unfamiliar technology falls sharply. We now choose what fits the job: .NET, TypeScript and React, Umbraco or Azure serverless.
Keep the tooling thin
We tried building elaborate custom tooling around AI and found the vendors shipped the same thing months later. We keep customisation light and put the effort into the work itself.
Adapt every week
The tools improve week by week, so our process does too. Something that didn't work in spring may work now.
Open use beats hidden use
We are open with clients about how we use AI, and we recommend they are open with their staff. Rules that people can follow work better than bans they work around.
Want to do the same?
We help small and mid-size businesses work out where AI fits, put sensible rules around it, and prove the value before scaling up.