18 years of business systems
We build the systems
a business actually runs on.
Delipat has spent eighteen years inside other companies' operations — their CRMs, their ledgers, their order books. On-premise AI is the newest layer we build. It is not the first, and that is exactly why ours works.
18+years building business systems
300+implementations delivered
12Salesforce certifications held
29client logos on our wall
Our story
From CRM engineering to AI that runs on your own hardware.
Delipat IT was founded in Durgapur, West Bengal, by Rajesh Chatterjee, a business-systems engineer who had spent years watching the same thing happen in company after company: good businesses running on software nobody had ever properly fitted to them.
We built the practice around fixing that. Salesforce, HubSpot, integrations into whatever the company already used. Over three hundred implementations later, across manufacturing, healthcare, retail, property and finance, in India, the United States, the United Kingdom and Australia, we know how a real operation holds together — and where it quietly falls apart.
Then generative AI arrived, and every business owner we spoke to wanted it. Almost none of them could use it. Their ledgers and customer records could not go into a public AI service, the per-message bill made no sense at scale, and the chatbots on offer could talk but could not do the actual work.
So we started building the other kind. AI that runs on hardware you own, reads your Tally and your CRM directly, and carries out real operational work — with a human approving every write before it touches live data. The AI part is new. The part where we make it work inside a real company is eighteen years old.
The person behind it
Meet Rajesh Chatterjee
Rajesh has spent eighteen years building the software businesses run on — first as an implementer, then as an architect, then leading a delivery team through more than three hundred projects for clients across India and the United States.
“Owners should not have to choose between using AI and keeping control of their own data. That is a false choice, and it is the one we removed.”
Today he writes the on-premise AI systems himself before anyone on the team touches them — the local models, the Tally sync, the approval gate that stops an agent writing to a live ledger. Every client engagement is one he is personally involved in. If you work with Delipat, you talk to him.
12 Salesforce certificationsSalesforce Ridge PartnerClickUp Platinum Partner300+ implementations
Talk to Rajesh
What drives us
Six rules we do not bend.
Rule 01Systems first, AI second
An AI is only as useful as the data it can reach. We fix the plumbing before we add intelligence, which is why our deployments survive contact with a real business.
Rule 02Nothing writes without a human
Every action an agent wants to take against live data waits in an approval queue for a named person. Built into the architecture, not offered as a setting.
Rule 03We tell you where the boundary is
You get it in writing: exactly what runs on your hardware, and exactly which parts touch anything outside. No claims we cannot stand behind in a security review.
Rule 04One department first
We would rather prove it in one team in six weeks than promise a company-wide transformation that stalls in month four. You widen it when you are ready.
Rule 05The numbers come from code
When you ask for outstanding receivables, the figure is calculated by software reading your ledger. The AI explains it. It does not invent it.
Rule 06Costs stay predictable
Running on your own hardware means a fixed monthly number instead of a bill that grows every time your team finds the system useful.
Who we build forBusinesses with real operations and real records.
Mid-sized companies where the owner still knows every number, and where losing control of the data is not an acceptable trade for convenience.
ManufacturingDistribution & tradingProfessional servicesHealthcare & life sciencesReal estateRetail & e-commerceLogisticsFinancial services
Come and look at your systems with us.
Thirty minutes, no deck. We look at what you run and what your team repeats every day, and tell you honestly whether on-premise AI is worth it for you yet.
