AI & intelligent automation
AI removes the busywork.
Judgment stays human.
Alchemy IBS Private Limited builds AI tools, prompt systems and workflow automation around a business's own repetitive work: enquiry replies, document processing, content drafts, knowledge search and data moving between systems. We build with OpenAI, Claude, Gemini, Zapier and Make. Every workflow that touches a customer or a consequential decision has a named person reviewing it.
We start with the task, then decide whether it needs AI.
Asking where to add AI produces demos. We ask which hour of your team's week looks the same every week. We watch the work first: how often it happens, and what it costs when it goes wrong.
Some of it needs a model. A lot of it needs a rule, and a rule is cheaper and more predictable.
Which kind of help does the work need?
| The work | What we build |
|---|---|
| Route an enquiry by product, region or source | A fixed rule in Zapier, Make or the CRM. No model needed |
| Answer the same first questions on WhatsApp or Instagram | A chatbot with scripted paths and a hand-off to a person |
| Draft content in your positioning and voice | Prompt systems built on your approved messaging, edited by a human |
| Find answers in manuals, policies or past proposals | Retrieval that cites the source document |
| Approve, price, diagnose or refuse | A person. We automate the paperwork around them |
Your positioning belongs inside the prompts.
Generic AI copy sounds like everyone else's. Because the team that found your idea also builds your AI tools, the prompt systems start from it: your approved claims, the claims you must never make, and the words your buyers use. The output still gets a human edit. It just starts much closer to your voice.
We didn't put a bot in front of buyers who wanted a person.
A US auto-parts business could have put a chatbot between its buyers and its sales desk. We didn't build one. Its buyers rang with a part number and a question, and a bot would only have stood between them and an answer. The useful work was making each call easy to place and easy to trace.
Where we draw the line.
Sensitive or high-accuracy workflows get human validation, audit trails, permission controls and a documented fallback. We don't build systems that make medical, financial or legal decisions unsupervised. We don't promise hours saved before a pilot has measured them.
So we start small: one task, agreed acceptance criteria, and a count of corrections and failures alongside the time saved. Expansion follows the evidence.
Name the hour your team repeats every week.
We'll tell you whether it needs a model, a rule or a person, and test one before building more. Most of our automation runs inside your CRM and follow-up, and when a workflow needs its own screens and users, it becomes a digital product.