AI agents that know your business - not the internet.
We build, ground, govern and operate custom AI agents on Microsoft Copilot Studio, Azure AI Foundry, Google Vertex AI and the open-source OpenClaw runtime. Trained on your documents, policies and data - so answers cite sources, not hallucinate them.
This is what a grounded agent feels like to users.
Every agent we deliver is grounded on your knowledge - SharePoint, Confluence, PDFs, ERP records, ticketing data - and refuses to answer outside that scope. Every reply cites the source document so users (and auditors) can verify the answer. Pick the runtime that fits: Microsoft, Google, or the open-source OpenClaw stack on your own server.
- Grounded on your own data with cited sources
- Source citation on every answer for audit & trust
- Tools / actions (raise ticket, post leave, lookup invoice)
- Multi-channel: Teams, web, WhatsApp, voice
- Guardrails - tone, topics, escalation rules
- Full audit log of every conversation
Knowledge sources
SharePoint - HR Policies Leave Handbook v4 Dataverse - Employees intranet.intrys.inActions / tools
Apply leave (Power Automate) Raise IT ticket Lookup payslipGuardrails
No salary info Escalate to HR after 2 retriesLive conversation
Platforms we build agents on
"We want AI - but not the kind that makes things up."
Off-the-shelf chatbots and unconfigured ChatGPT prompts don't know your business. Worse, they confidently invent answers. Grounded agents fix that.
- "HR answers the same 20 questions every week."
- "Support agents copy-paste from a 200-page KB nobody reads."
- "Sales reps can't find the right deck for the right industry."
- "Parents call the school front desk for fees, timetable, holidays."
- "We tried ChatGPT - but it made up our refund policy."
What every agent build includes
- Use-case workshop & conversation design
- Knowledge ingestion (SharePoint, PDF, web, DB)
- Grounded retrieval pipeline (embeddings + vector store, or OpenClaw skills)
- System prompt, persona & guardrails
- Tools / actions (Power Automate / REST APIs / OpenClaw plugins)
- Channel deployment (Teams / web / WhatsApp / voice)
- Evaluation set (50 Q&A pairs scored before go-live)
- Audit log + content-safety filters
- Admin handover + 30-day post-launch tuning
Built for the work that quietly slows your team down.
Don’t see yours? Tell usEvery Intrys agent is built around a real, repeating job — not a demo. Browse 12 production-ready use cases across the three teams that ask us for AI most: accounting & finance, sales & quoting, and business operations.
Invoice & GST query agent
Instant, cited answers on POs, invoices, GSTR-2B reconciliation and input credits — grounded on Tally, Zoho Books and GSTN portal exports.
Tally / Zoho Books assistant
Natural-language access to ledgers, trial balance and ageing reports — no more waiting for someone to “pull a report.”
Compliance & filings agent
Tracks GST, TDS, PF, PT and ROC due dates against your filings, flags gaps and drafts reminder mails to the accountant or auditor.
Expense & reimbursement bot
Submit, query and approve expense claims from Teams or WhatsApp. Reads receipt photos, auto-extracts vendor, amount and GST.
Instant quote-creation agent
“Quote me 25 Dell laptops with 3-yr ProSupport + M365 Business Standard.” Pulls latest pricing, applies your margin & discount rules, drafts a branded PDF in seconds.
Proposal & deck builder
Generates a tailored proposal deck from your library — right case studies, industry slides, pricing tables, cover letter — branded and ready to send.
Pricing & SKU lookup agent
“Latest MRP on HP EliteBook 840 G11? Our CSP discount on Azure?” Always-current prices, GST clarity, partner-only deals — with sources.
Lead qualification & first-reply agent
Greets inbound website / WhatsApp leads, asks 3–4 qualifying questions, books a Calendly slot and posts a summary to your CRM.
Executive briefing agent
Every morning at 8am, a one-screen summary: revenue MTD, top 5 risks, overdue tasks, customer escalations — grounded on Power BI, CRM and email.
Project status & RAID agent
“How is the Acme rollout tracking? Any blockers this week?” Reads Planner / Jira / project SharePoint and rolls up RAG status, RAID and next steps.
Meeting notes & action agent
Joins your Teams calls, takes notes, drafts minutes, extracts action items with owners & due dates and posts them to Planner / email.
Vendor & PO management agent
“Which POs are pending GRN? Which vendor missed 3 SLAs in a row?” Reads your ERP / Tally / vendor folder and answers with source links.
Three Intrys agents, at work.
Pixel-accurate previews of how the agents above look in real life. Same UI you’ll get — on Teams, WhatsApp or your own website.
Top overdue (₹):
1. Acme Industries — ₹ 18.2 L (62 days)
2. Bharti Logistics — ₹ 11.7 L (48 days)
3. Cloud9 Schools — ₹ 7.4 L (41 days)
4. Veritas Pvt Ltd — ₹ 5.9 L (39 days)
5. Sundar Steel — ₹ 4.2 L (37 days)
25 × Latitude 5550 — ₹ 74,200 each
+ 3-yr ProSupport — ₹ 6,800 each
+ M365 Business Std — ₹ 770 / user / mo
Subtotal: ₹ 20,42,500
Edu discount (10%): –₹ 2,04,250
GST 18%: ₹ 3,30,885
Grand total: ₹ 21,69,135
• Revenue MTD: ₹ 1.84 Cr (89% of target) ▲
• Pipeline coverage Jun: 2.3× (healthy)
• 🔥 Escalation: Acme — Wi-Fi outage, P1 open 2h
• 📅 Today: 4 demos, 2 SOWs to sign, board pre-read by 6pm
• ⚠ Overdue: 3 invoices, 1 PF filing (due tomorrow)
What you actually save in the first 90 days.
Conservative numbers from real Intrys deployments in Indian SMBs (50–500 staff). Your mileage depends on volume — we forecast yours in the workshop.
Platforms & channels we build on
Not sure which? Ask usMicrosoft stack
Copilot Studio
Low-code agents grounded on SharePoint, Dataverse, web & PDF. Best for HR, IT helpdesk, FAQ bots.
M365 Copilot agents
Custom agents living inside Microsoft 365 Copilot for your power users (Word, Teams, Outlook).
Azure AI Foundry
Pro-code agents with GPT-4o, custom tools, multi-agent orchestration & observability.
Azure OpenAI Service
GPT-4o / o-series / embeddings with EU/India data residency & enterprise SLA.
Google & open-source
Google Vertex AI
Gemini 2.x family + agent builder for Google Workspace shops & multi-cloud setups.
OpenClaw runtime
Open-source personal / team agent runtime that runs on your infra (Mac, Linux, Windows, even a Raspberry Pi). Chat to it from WhatsApp, Telegram, Slack, Signal, Discord or iMessage. Multi-model (Anthropic / OpenAI / local), persistent memory, browser & system access, hot-loadable skills via ClawHub - your data never leaves your box.
Channels we deploy to
Microsoft Teams
Personal & channel agents, adaptive cards, single-sign-on with Entra ID.
WhatsApp Business
Customer-facing agents for orders, support, school updates, appointment booking.
Website widget
Drop-in chat widget for your marketing site or product app - dark/light themed.
Voice / IVR
Azure Speech + agent backend for inbound helpdesk / school front-desk lines.
How we build an agent - live in 3-6 weeks
1. Discovery
Half-day workshop. We pick the use-case, define personas and write a 50-question evaluation set.
2. Knowledge
Ingest your docs, chunk & embed, build the vector store, test retrieval quality.
3. Agent build
System prompt, persona, tools / actions, guardrails, escalation paths.
4. Evaluate
Score answers against the eval set. Iterate prompts & chunks until quality threshold is met.
5. Deploy & tune
Publish to Teams / web / WhatsApp. Monitor conversations. Tune weekly during hypercare.
Agent build packages
Need a custom mix? Ask usSmall business agent
₹ 65,000
Grounded Q&A on your docs- 1 agent on Copilot Studio
- Up to 200 documents / 1 SharePoint site
- Cited answers + guardrails
- 1 channel (Teams or web)
- 50-question evaluation set
- 30-day post-launch tuning
Business agent
₹ 85,000
Knowledge + actions + multi-channel- 1 agent on Copilot Studio or Azure AI Foundry
- Up to 2,000 docs across 3 sources
- 3 tools / actions via Power Automate
- 2 channels (Teams + web or WhatsApp)
- Persona, brand tone, guardrails
- 100-question eval set + bias / safety review
- 60-day post-launch tuning
Custom grounded agent
From ₹ 4,25,000
Azure AI Foundry / Vertex AI / OpenClaw- Pro-code agent on Foundry / Vertex / OpenAI
- Custom grounded retrieval + vector DB
- Multi-tool orchestration (5+ tools)
- Multi-channel + voice / IVR option
- Observability dashboards + cost monitor
- CI/CD + eval harness in your repo
- 90-day post-launch tuning
OpenClaw on-prem agent
₹ 85,000
Your runtime, your data, your rules- OpenClaw installed on your server / Mac / Linux box
- Bring-your-own model keys (Anthropic, OpenAI, local)
- 2 channels (WhatsApp / Telegram / Slack / Signal / iMessage / Discord)
- 3 starter skills from ClawHub or custom
- Persistent memory + system prompt tuned for your team
- Handover + 30-day tuning
Multi-agent & managed
Custom
Several agents + governance- 3+ agents (HR, IT, sales, customer)
- Shared knowledge + tool library
- Responsible AI policy & sign-off
- Eval harness with regression tests
- Managed agent ops (monthly)
- Usage & cost reporting
All prices are in INR, excluding 18% GST. Microsoft / Azure / Google subscription & token usage billed separately - see licence table below.
Common add-ons
Need something else? Ask| Add-on | What you get | Price (INR, ex-GST) |
|---|---|---|
| Extra channel | Deploy an existing agent to another channel (Teams / web / WhatsApp / Telegram / voice). | From ₹ 22,000 |
| Extra knowledge source or tool | New connector (SharePoint / DB / API) or new action (Power Automate / REST / OpenClaw skill). | From ₹ 22,000 |
| WhatsApp / Voice setup | Meta Business verification & templates, or Azure Speech + telephony + agent backend. | From ₹ 32,000 |
| Responsible AI review | Bias, safety, privacy & hallucination audit with a written report. | From ₹ 55,000 |
| Maker training (2 days) | Hands-on workshop on Copilot Studio / Foundry / OpenClaw for your team. | ₹ 55,000 |
Managed agent operations - monthly retainer
| Plan | Included monthly | SLA | Price (INR, ex-GST) |
|---|---|---|---|
| Agent Care | Conversation review, knowledge refresh (1 source), 4 small tuning changes, usage report. | Next-business-day | ₹ 22,000 / mo |
| Agent Growth | Care + 12 hr of new tools / channels / prompt work, eval re-run, cost tuning. | 4 business hours | ₹ 48,000 / mo |
| Agent Factory Ops | Dedicated AI engineer (40 hr/mo), multi-agent governance, RAI sign-off, eval harness. | Same-day | From ₹ 1,15,000 / mo |
Unused hours roll over for one month. Annual contracts get one month free.
Platform licences & usage
Procure via Intrys CSP| Licence / usage | For whom | MRP (INR) | Notes |
|---|---|---|---|
| Copilot Studio | Each tenant publishing custom agents to Teams / web. | ₹ 16,800 / mo | 25,000 messages / month included; overages billed at packs of additional messages. |
| Copilot Studio messages | Overage on top of base pack. | ₹ 8,400 / 25K msgs | Each generative or action turn typically counts as 1 message. |
| M365 Copilot | Power users who want Copilot in Word / Teams / Outlook + custom agents. | ₹ 2,500 / user / mo | Annual commit; agents extend the same M365 Copilot UX. |
| Azure OpenAI - GPT-4o | Pro-code agents needing strong reasoning. | ~₹ 220 / 1M input tokens | Usage-based. ~₹ 880 per 1M output tokens. India South / UAE region available. |
| Azure OpenAI - GPT-4o mini | High-volume, low-latency agents (HR, FAQ). | ~₹ 13 / 1M input tokens | Typically 80-90% cheaper than GPT-4o for similar quality on RAG. |
| Azure AI Search | Vector store for RAG grounding. | From ₹ 6,800 / mo | Basic tier is enough for <50K docs. |
| Vertex AI - Gemini | Google Workspace customers or multi-cloud setups. | Usage-based | Gemini 2.x Pro / Flash pricing per 1M tokens; passes through Google MRP. |
Microsoft & Google prices change periodically. Always quoted at current MRP at order time. CSP customers may receive promotional discounts. Token figures rounded for illustration - your proposal will use exact, current rates.
Fixed scope, fixed price - usage passed through.
The build packages above are fixed-scope, fixed-price. Token / message usage is billed at MRP directly to you (or via Intrys CSP), so there's no markup on the AI bill itself. If your project doesn't fit a tier, we quote T&M at ₹ 16,000 / engineer-day with a written cap.
- Free 30-min AI workshop before any quote
- 50% advance, 50% on go-live for build work
- Usage billed monthly at platform MRP
- GST invoice, MSME-registered vendor
- Annual managed contracts get one month free
Our guarantees
- Grounded or refund - if we don't hit 85% accuracy on your eval set, you don't pay the build fee.
- Your data, your tenant - all knowledge stays in your Microsoft / Azure / Google tenant.
- No training on your data - Azure OpenAI & Vertex don't train on prompts by default; we keep it that way.
- Documented prompts & eval set - you can take it anywhere.
- Exit clause - clean handover at any month-end.