Weeks 1-2
Frame
The outcome written as a number, not an ambition
Baseline measured against production data
Kill criteria agreed before a line is written
Enablement & Assurance · India
Four on paper. Forty in practice.
Most companies declare four AI systems. We find the rest in week one, then make each defensible, owned and measured.
The gap isn't ambition. It's execution.
The failure modes are consistent and they are documented: poor data quality, legacy systems that won't integrate, weak governance, and objectives nobody wrote down. None of them are model problems. All of them are solvable.
56%
of companies report no significant financial benefit from their AI investments.
BCG, AI Radar, 2024
42%
of enterprises abandoned most of their AI initiatives during 2025.
S&P Global Market Intelligence, 2025
27%
of the average Indian mid-market AI budget is lost to complexity overhead.
72 Networks analysis, methodology to publish
₹33,000 cr
written off against stalled or abandoned AI programmes in India, every year.
72 Networks estimate, methodology to publish
Three things, done properly.
Assure
AI Governance & Assurance
Deploy AI you can defend.
- Six to ten weeks
- Board-ready assurance report
- Surveillance retainer, monthly
Operate
Managed AI Ops + Fractional CAIO
AI that is still working in month twelve.
- Managed operations, per system
- Fractional CAIO, two to four days a month
- 90-day notice, no lock-in
Build
Function Pods & Pilot Rescue
Eight to twelve weeks, a named outcome, a measured baseline.
- Eight to twelve weeks
- Finance, Customer, Sales, Knowledge
- 70% fixed, 30% on the metric
Four gates. Twelve weeks.
Every engagement passes the same four gates in the same order. Each one ends in something written down that you can hold us to.
Weeks 3-5
Ground
Data readiness verified against production data, not a clean extract
Evaluation harness and golden dataset built before any model is touched
Integration surface mapped and owned
Weeks 6-10
Build
Shipped against the harness, not a demo script
Governance artefacts produced as you go, not retrofitted
Your team in the room, on the tools
Weeks 11-12
Land
Handover with a measured capability score
Metric verified ninety days after go-live
Professional indemnity cover on every outcome clause
Thirty percent of our fee rides on your number.
- 70%
- Fixed. 40% on signature, the balance across three-week milestones
- 30%
- On the measured metric, ninety days after go-live
The metric is agreed before we start, measured against a baseline we take in week one, and capped at 30% of the total fee. If we miss it, we don't earn it.
What we don't do
- Rent out engineers by the month. We'll refer you to someone who does that well.
- Train foundation models. Almost never right for a company your size.
- Take work we don't believe we can land.
- Sell licences, or take vendor referral fees.
- Stay longer than the handover needs. Our contract requires us to leave.
Four sectors, known well.
Banking, NBFCs, Insurance & Capital Markets
You carry the liability. Even when someone else built the model.
Manufacturing
Quality, maintenance and planning AI on the shop floor, not in a slide deck.
Global Capability Centres
AI delivery your global parent can audit.
Professional Services
Research, drafting and knowledge work, measured against the fee earner's clock.
The way in
The 30-Day AI Reality Check.
Four weeks. We map every AI system you own, measure what it costs against what it returns, and hand you a plan with owners and dates. Fixed fee, credited in full against any engagement that follows.
- Fixed
- Fee
- 30 days
- Start to board pack
- 4 a month
- Engagements taken
- 100%
- Credited against what follows
Research, not blog posts.
Every piece has a named author, a date, and a source list. Anonymous thought leadership from an unknown firm converts at approximately zero. Ours is signed.
Briefing · Regulation
SEBI Regulation 16C: what “solely responsible” means in practice
Legal can tell you that you are liable; only operations can tell you what you are liable for.
Meera Krishnan · 12 Aug 2026Field notes · Delivery
What forty AI inventories taught us about shadow AI
The HR tool that started scoring résumés in a March release note nobody read.
Rohan Iyer · 4 Aug 2026Research · BenchmarkThe Mid-Market AI Waste Index, Q3 2026
What 214 self-reported AI portfolios say about where the money actually goes.
Meera Krishnan · 28 Jul 2026