The questions you should ask any AI firm. Our answers.

You should not take it on trust, so we make it easy to check. We start small: a fixed-price Sprint or pilot, scored against a metric and an evaluation set we agree in writing before we build. You see the results on your own data before you commit to more. We have no client logos to show you yet, and we will not invent any. What we can show you is how we think: our articles, our method and a detailed proposal for your problem.

Four things. A named owner who can make decisions. Two to four hours a week from the people who do the work today, so we understand the real workflow. Access to sample data and to the systems we will connect to, under an NDA and your security rules. And someone from IT or your IT vendor for integration and security review. We will send a full checklist before the engagement starts.

For most systems, 8 to 17 weeks from the first workshop. The Sprint takes 2 to 3 weeks, the pilot 4 to 8 weeks and the production build usually 2 to 6 weeks more. Narrow, well-defined work such as a single reconciliation can go faster. Work that touches regulated decisions or many legacy systems can take longer. We give you a dated plan in the proposal.

Fixed price per phase, agreed in writing before each phase starts. 40% in advance, the rest on milestones, each due within 15 days of invoice. Managed AI is a monthly retainer with a six-month minimum term. Model, API and cloud costs are passed through at cost plus a 10% handling margin when we hold the accounts (no margin if you pay the providers directly), on a separate line, never absorbed. Indicative ranges for every service are on our pricing page.

We work inside your environment wherever we can, so data stays where it already lives. Access is least privilege, time-limited and logged. Personal data such as Aadhaar, PAN, card numbers and phone numbers is redacted before it reaches a model or a log where the use case allows. Every system ships with access controls, audit logs, guardrails against prompt injection and a security review before go-live. At the end of an engagement, we delete any copies we hold and confirm it in writing.

By default, in India. We prefer to deploy in your own cloud account, in an Indian region, or on your own servers. If a commercial model API is the best option, we use its India region where one exists and tell you in writing where each piece of data goes. If your regulator or your policy requires data to stay in India, we design for that from the start.

No. Your data is never used to train our models or anyone else’s. We only use model providers on terms that exclude training on customer data, and we confirm the retention terms for the specific product tier we use. Evaluation sets built from your data belong to you and stay in your environment.

We design every system to meet the Digital Personal Data Protection Act 2023 and the DPDP Rules 2025. Most core duties apply from 13 May 2027, and we build to them now. That means clear notice and consent, data minimisation, retention limits, breach readiness and support for data principal requests. Compliance is ultimately yours as the data fiduciary, so our AI Governance and DPDP Readiness service helps you set the policies. It is practical guidance, not legal advice; your lawyer should review it.

Yes. Financial services is one of our four focus industries. We design for data localisation, outsourcing guidelines, audit trails and human oversight of decisions that affect customers. We work with your compliance and information security teams from the first week, not at launch. We do not give regulatory opinions; we build to the requirements your compliance team confirms.

The one that scores best on your evaluation set, at a cost and speed that works, within your data rules. That may be a commercial model API or an open-weight model hosted in your environment. For Indian languages, we test Indic models alongside global ones. We design systems so the model can be swapped later without a rebuild, because prices and quality change every few months.

No. We are not an official partner or reseller of any model or cloud provider. That is deliberate. It means we have no reason to recommend one over another except that it works better for you.

Yes, where the use case allows. Open-weight models can run on your own servers or in a private cloud, with no data leaving your network. The trade-offs are hardware cost and, for some tasks, lower accuracy than the largest commercial models. We measure both during the pilot so you decide with numbers.

We build for English, Hindi, Marathi and Hinglish first, and test other Indian languages such as Tamil, Telugu, Kannada, Bengali and Gujarati per project. Quality varies by language, task and channel, especially on voice. So we do not promise a language until we have tested it on your real data and shown you the scores.

Sometimes you should, and we will tell you. A SaaS tool is right when your process is standard and the tool already connects to your systems. We are useful when the work depends on your own systems, such as Tally, SAP, a core banking platform or an old database, when you need Indian languages done well, when data must stay in India, or when you need measured accuracy before customers see it.

You own what we build for you: the code, prompts, configurations, evaluation sets and documentation. We keep the tools and libraries we had before working with you, and you get a perpetual licence to use them inside the delivered system. The exact terms are in the contract, which your lawyer should review.

Then you know, for a fixed price, before spending on a full build. The pilot is scored against a success metric agreed in writing. If it misses, you get the full evaluation report, the reasons and our honest recommendation, which may be to stop. You are under no obligation to continue.

Two parts. The Managed AI retainer, from ₹75,000 to ₹5 lakh per month excluding GST, covers monitoring, weekly quality checks, improvements and support. Then running costs: model, telephony and cloud usage, passed through at cost plus a 10% handling margin when we hold the accounts (no margin if you pay the providers directly). We measure the cost per conversation, call or document during the pilot and give you an estimate in writing before go-live.

Yes, before we see any of your data. We are happy to sign your NDA or send our mutual one. Confidentiality and data handling terms are also written into every engagement contract.

Our own engineers, working with your team, on site when it helps and remotely when it does not. A founder is involved in every engagement. We do not subcontract the core build without telling you in advance. And yes, we work with your IT vendor or system integrator. They know your systems; we bring the AI. We agree a clear split of responsibilities at the start and keep them in every review.

You keep everything. Before the retainer ends, we hand over the code, prompts, evaluation sets, dashboards, runbooks and access, and walk your team or your vendor through how to run and improve the system. Nothing we build depends on an Envisstack account you cannot replace. The retainer has a three-month minimum, then 30 days’ notice.