CoreConsultAI in software development – working outline
Since late 2022, when ChatGPT first arrived in the hands of mainstream developers, AI has been quietly but steadily changing how we build software. What started as “paste a snippet, get an answer” has turned into a set of tools that sit in the editor, in CI, in chat – and in practice touch almost every step of the delivery pipeline.
This article itself is a good example: the structure and large parts of the wording were drafted with the help of AI, then reviewed and adapted by humans. That combination – AI as a fast, tireless assistant and engineers as editors, decision‑makers and owners of quality – is exactly how we like to work.
From ChatGPT in the browser to AI everywhere
- 2022–2023: ChatGPT in a tab – first experiments, ad‑hoc questions, code explanations
- Next step: AI in the IDE – inline suggestions, quick refactors, test scaffolding
- Today: multiple AIs in parallel – we use ChatGPT, Claude, Grok and Gemini side‑by‑side, choosing the one that fits the task or cross‑checking answers when something is critical.
Instead of a single “magic” model we treat AI tools like we treat libraries: each has strengths and trade‑offs, and we combine them to get the best outcome for a given problem.
Where AI fits into our delivery pipeline
- Discovery and design – exploring solution options, generating UI/UX sketches, drafting acceptance criteria and edge‑case lists.
- Implementation – writing boilerplate, suggesting alternative APIs, translating legacy patterns into modern ones (for example moving from ad‑hoc React to a Next.js + design‑system setup).
- Testing – generating realistic test data, proposing e2e scenarios for Playwright, helping with fixtures and mocking strategies.
- Maintenance – summarising large PRs, explaining unfamiliar code paths, suggesting refactors that reduce complexity.
The goal is not to hand over decisions to a model, but to use AI as a force multiplier: less time on boilerplate and yak‑shaving, more time on product thinking and real trade‑offs.
Productivity and the T‑shaped to full‑stack shift
One of the strongest effects we see is how AI reshapes individual roles. Classic T‑shaped developers – deep in one area, reasonably broad in others – can now move much faster along the horizontal bar of the “T”. With AI helping to fill gaps in unfamiliar frameworks, languages or tooling, it becomes much more realistic to work as a full‑stack engineer on a day‑to‑day basis.
Concretely this means: backend‑heavy people can navigate modern frontend stacks (React, Next.js, TypeScript, Playwright) with less friction, and frontend‑heavy people can be more productive around APIs, infrastructure and data. The result for the organisation is higher flexibility in staffing, fewer hand‑offs between silos and faster end‑to‑end delivery of features.
Risks, guardrails and human responsibility
The productivity gains are real, but so are the risks: hallucinated facts, subtle bugs, security oversights and code that “looks right” but does the wrong thing. Our stance is simple: AI may write a lot of the draft, but humans remain accountable for architecture, correctness, security and ethics.
- We review AI‑generated changes like any other code.
- We rely on tests (unit, integration, e2e with Playwright) and static analysis, not on trust.
- We keep domain knowledge and key decisions inside the team, not inside prompts.
Used this way, AI is less a replacement and more a lever: it lets us ship higher‑quality software faster, with smaller teams – and it makes the work itself a bit more fun.