Application candidate

Fixed rules score this career assessment.
AI only puts the result into words.

Alchemy IBS Private Limited built a career-assessment application in which fixed scoring rules calculate the result and AI only explains it. The obvious design lets a language model read the answers and pronounce a career. We separated the two, so the same answers always produce the same score, and the report leads to a conversation rather than a verdict.

Book a strategy conversation

At a glance.

Client	Industry	Market	What we did	Stage
A career-guidance product	Education and careers	India	Product design, deterministic scoring, integrity checks, bounded AI explanation, follow-up journey	Application candidate

People had to leave the assessment knowing more.

The product helps a person reflect on their career direction and decide what to do next. Questions go in, a result comes out, and the person should leave knowing more than when they started.

The build most teams would reach for in 2026 is simple to describe: send the answers to a model and let it write the career advice.

People treat a career result as authoritative. So the result has to be checkable.

A person reads a career report as a judgement about their future, however it was produced. If generated prose decides the result, two people with identical answers can be told different things, and nobody can show why either was told what they were.

We gave each part one job. Rules calculate. The model explains, within limits. The report opens a next step. The score comes from defined rules applied the same way every time, which means it can be tested against known inputs. The model's job is to put that result into plain language using only the context it is given, and its output can be checked for claims it should not make.

The third part mattered as much as the first two. A score on its own leaves a person alone with a number. Ending the report in a career conversation turns the assessment into the start of guidance, not a substitute for it.

Five layers, and what each is allowed to do.

Layer	Its job
Structured questions	Capture answers in a staged flow, so the person is never faced with everything at once
Deterministic scoring	Apply defined rules consistently, testable against known inputs
Integrity checks	Flag response patterns that need care before a result is shown
Bounded explanation	Describe the result in plain language, within the supplied context
Follow-up journey	Carry the person from the report into a further career conversation

A working candidate, run locally and not yet released.

The application is a working candidate, built and run locally, and not yet released. It is a tool for career reflection and guidance. It is not a validated psychometric test or a predictor of career success, and the report is written so it never sounds like one.

Before AI goes into a product that judges people, decide what it may decide.

Let rules hold the judgement and let the model explain it.

Explore education · See digital products · Book a strategy conversation