We build AI that does real work, and proves it.
8 agents in productionDubai · PuneBacked by T-Hub
We kept watching AI stop one step short of the work.
So we built the thing we kept wishing existed.
We met inside Deutsche Telekom, building AI into regulated workflows. The demos were impressive. Then a person copied the model's answer and typed it into SAP by hand. In 2023 we left to build agents that finish the job: read the document, make the decision, post it into the system of record.
Meet the agentsThe pattern
Years inside a regulated enterprise watching AI stop short of execution. Great answers, typed into SAP by hand.
The decision
In 2023 we founded Attentions AI to build agents that finish the job, on your hardware, with your data.
In production
Started in finance, automotive, healthcare, and engineering. Those agents now run every day on real company data.
Four things we won't compromise on.
- AI should earn its keep. If it costs more to run than the work it replaces, it isn't worth it.
- Your data is yours. It shouldn't have to leave your walls for AI to be useful.
- An answer you can't check is an answer you can't trust.
- Production in weeks beats a pilot that drifts for months.
From a short assessment to production in weeks.
We start with one workflow, prove it on your own data, and put it live. The risk sits with us, not you.
- Step 1
Assess
A fixed-fee assessment. We sit with your team, look at your real workflows, and pick the first agent to build. If it isn't a genuine fit, you've spent very little to find that out.
- Step 2
Build
We assemble the agent from parts we've already proven elsewhere, so the build stays short and the timeline holds.
- Step 3
Run and scale
Once the first agent is working and trusted, most customers don't stop at one. The next one reuses the same parts.
Four commitments, in writing
Fixed-fee assessment
An honest go / no-go first.
2-4 weeks to production
Real work, not a pilot.
Runs on your infrastructure
On-prem or your own cloud.
Payback within 6 months
Our written promise.
Where the agents run.
Each one doing a real job today
Finance to healthcare to engineering
Dubai · Pune
Backed by T-Hub's AI incubation
Finance
Invoice & voucher matching.
Automotive
Defect & warranty intelligence.
Healthcare
Clinical notes & patient access.
Engineering
Tender & BOQ intelligence.
Who does the work
Built by the people on the call.
Two founders and three senior leaders, plus 18 team members across AI research, ML engineering, data science, and domain expertise.
We met building AI into regulated enterprise workflows, and left in 2023 to build agents that finish the job.
Read the document, make the decision, post it into the system of record, on your own hardware, where your auditors can check every step.
Finance, automotive, healthcare, engineering: the workflows with real money and real rules attached. Those agents run every day.
+188 agents in productionFounders who shipped. Advisors who've scaled.
Founder & CEO · Dubai
Puneet Kumar Ojha
Spent years inside Deutsche Telekom as an Enterprise Architect, implementing AI across regulated workflows and watching it stop short of execution. The pattern repeated everywhere: impressive demos, brittle production, and teams still typing AI output into SAP by hand. Founded Attentions AI in 2023 to build the platform that finishes the job, on your hardware, with your data, for your regulator.
Founder & CTO · Pune
Ankit Agrahari
Architected enterprise integrations at Deutsche Telekom, the deep system-integration work that makes artiGen's execution layer possible. Knows what breaks when an AI output tries to post itself into a live ERP, and how to architect around it. Leads 18 top experts in AI, ML, product engineering, and domain expertise.
Advisory board
Omair Ahmed Khan
EVP of Data, Analytics & AI
Swisscom
Drives enterprise transformation with deep expertise in AI strategy, data governance, and large-scale change management across regulated industries.
Monica Dalla Riva
Chief Design Officer
Deutsche Telekom
Shapes customer-focused innovation through global design and product expertise. Two decades architecting digital products at enterprise scale.
Murad Semercioglu
CEO
CPC Yönetim Danışmanlık
Guides leaders and organizations through strategic transformation with expertise in executive advisory and performance coaching.
Plus 18 team members across AI research, ML engineering, data science, and domain expertise.
The partners we build with.
BytePlus
BytePlus, a ByteDance company, is one of the model platforms we work with. By default we tune an open-source small language model to your workflow; when you'd rather use your own or a specific model, BytePlus is one of the options we plug in. The agents and the guardrails are ours either way.
T-Hub
We're part of T-Hub's AI incubation, which gives us the R&D and compute to build specialised, sovereign AI agents.
Want one of these agents on your workflows?
Start with a short assessment. We'll look at one workflow and tell you honestly whether it's a good fit.