Grow the top line
Faster quotes and proposals, better follow-up on leads, offers matched to each customer.
Fractional Head of AI · for mid-sized companies
AI that shows up in your P&L.
Senior AI leadership for companies that need it but aren't ready for a full-time hire. I find where AI pays off in your business, prove it with small, measured pilots, and train your people to run it. Manufacturing, services, hospitality, distribution and beyond.
Twenty years building data and AI systems across seven industries, from banking and insurance to healthcare and e-commerce. Now working directly with growing businesses.
1–3 days
a week of senior AI leadership, at a fraction of a full-time hire
4 weeks
from first meeting to a board-ready AI roadmap
90 days
to two or three pilots running against a measured baseline
12 months
to five or more use cases scaled and your teams trained
Where most companies are stuck
You know AI matters. The hard part is knowing where to start.
- Everyone talks about AI. Nobody has said where it pays.Vendors pitch tools. Nobody has mapped your quotes, orders, inspections or bookings against what AI can actually do today.
- A senior AI hire is hard to find and hard to justify this early.You need judgment at leadership level before you know how big the programme should be.
- Pilots start with excitement and end without a number.No baseline, no success metric, so no decision on whether to scale it or stop.
- Customer data, drawings and IP feel exposed.Staff are already pasting work into public tools, with no policy on what is allowed.
- Your people are uneasy.Some worry about their jobs, others are quietly experimenting. Neither group has been brought into a plan.
What you should expect to see
Every use case has to move one of these.
We agree the measure before any work starts. If a project can't show movement on one of these within its pilot window, we stop it.
Protect the margin
Fewer manual hours on repetitive work, less rework, less spend on outsourced back-office tasks.
Raise quality
Fewer defects reaching customers, consistent answers from the frontline, cleaner documentation.
Shorten cycle time
From enquiry to quote, order to dispatch, ticket to resolution. Measured before and after.
Stand out in your market
A confident, specific answer when customers ask how you use AI, backed by policy and results.
Bring the team along
Role-specific training, AI champions in every function, and adoption that lasts after I step back.
When AI is the wrong tool for a problem, I'll say so and point you to the simpler fix. A spreadsheet, a process change or an existing feature in software you already pay for is often the better answer.
How the engagement runs
From first conversation to AI as the way you work.
Four phases with a clear exit at each one. You decide whether to continue at every step, based on results you can see.
1
Weeks 1 – 4
Diagnose
Interviews with leadership and every function head. A look at your workflows, systems, current AI use, tool spend and data.
You get
- A ranked use-case pipeline, by impact and effort
- A 12-month roadmap with owners and measures
- A draft AI usage and data policy
2
Months 2 – 3
Prove
Two or three time-boxed pilots, each with a baseline, a hypothesis and one success metric. Build-or-buy decided per case.
By day 90
- Pilots running with real users
- First results against the baseline
- Policy reviewed and in use
3
Months 4 – 12
Scale
Proven pilots roll out beyond their first team. Training by role, playbooks, prompt libraries and a champions network in each function.
By month 12
- Five or more use cases proven and scaled
- Measurable gains in your top three workflows
- Most eligible staff using approved tools weekly
4
Months 12 – 24
Hand over
Function heads own their AI workflows. My role shifts to the harder problems and a lighter advisory rhythm, until you don't need me.
The end state
- AI-assisted work is the default
- Champions sustain adoption on their own
- Zero data or IP incidents
Every idea passes the same gates
A shared pipeline keeps leadership honest about what is working. Weak ideas are retired early, with the reason written down.
Where AI tends to pay first
A starting map for your industry.
These are common first candidates, plotted by likely impact and effort. Your discovery phase replaces this with a map built from your own workflows and numbers.
Positions are indicative. Tap a point or a row to link them.
What I take on
One person across strategy, build and adoption.
Big consultancies split these across separate teams. For a mid-sized company, having one accountable person across all six is usually faster and cheaper.
AI strategy & roadmap
A company-wide view of where AI creates leverage, function by function, ranked by impact against effort.
- Rolling 12-month roadmap
- Investment cases for leadership
- Quarterly progress reviews
Use-case discovery
Structured sessions with sales, operations, quality, finance, HR and support to find high-volume, high-friction work.
- Live use-case pipeline
- Clear no to hype-driven ideas
- Build-versus-buy calls
Pilots & measurement
Time-boxed pilots with a baseline and one success metric. Results reported honestly, including where AI didn't help.
- Cycle time, effort, cost per output
- Defect and error rates
- Scale-or-stop decisions
Hands-on build
I still build. Assistants over your manuals and documents, workflow automation, and AI agents connected to the systems you already run.
- ERP, CRM, ticketing, email, shared drives
- Private models on your own hardware
- Search over drawings, SOPs and records
Team enablement
Training by role, not generic webinars. Playbooks, prompt libraries, office hours and a champions network inside each function.
- Leadership AI workshops
- Frontline and shop-floor training
- Open conversation about job impact
Governance, security & IP
A usage policy covering customer data, designs, source code and NDAs. Approved tools, data rules and review gates.
- Fits ISO and customer audits
- Your answer when clients ask about AI
- Audit trails from day one
Ways to work together
Start small. Commit only when you see results.
4 weeks · fixed fee
AI Readiness Sprint
For leadership teams who want a clear, costed plan before committing budget.
- Baseline of AI use, tool spend and data readiness across functions
- Ranked use-case pipeline with impact and effort
- 12-month roadmap with owners and measures
- Draft AI usage and data-handling policy
- Two or three pilot designs, ready to run
You walk away withA plan your board can approve, whether or not we continue.
6 – 12 months · 1 – 3 days a week
Fractional Head of AI
For companies ready to act that need senior leadership to run the programme.
- I own the roadmap, pilots and build-versus-buy calls
- Monthly leadership reviews with honest metrics
- Enablement programme and champions network
- Governance and policy in force
- Hands-on build where it speeds things up
You walk away withScaled use cases, a trained team and an internal owner ready to take over.
8 – 12 weeks · per pilot
Pilot & Enable
For teams with one specific problem worth solving now.
- One time-boxed pilot, from baseline to result
- A working solution connected to your systems
- Role-specific training and a playbook
- A scale-or-stop decision backed by numbers
You walk away withA proven workflow in daily use, or a clear reason it shouldn't be.
Also available: a one-day AI workshop for founders and management teams. Scope and price are agreed in writing before any engagement begins.
Built, not just advised
Where the experience comes from.
I've spent two decades inside the systems I now advise on: designing them, building them, running them in production and leading the teams behind them.
Renewable energy · AI platform · 2024 – present
An AI operations platform, built from zero
As founding technical leader, I architected an AI platform that reads plant telemetry, maintenance history and equipment manuals, then produces diagnostics, forecasts and prioritised maintenance actions. It turned static electrical drawings into a knowledge graph the system can reason over.
~80%less engineering analysis time0 → 20engineers hired and led
Why I start with the documents and drawings you already have.
E-commerce · Global marketplace · 2020 – 2024
The data platform behind company-wide decisions
Led the global team running the analytics warehouse: 700+ daily pipelines from 300+ services, feeding reporting, marketing and machine learning. Led two back-to-back cloud migrations and presented every trade-off to senior leadership.
~30%lower infrastructure cost<5 hrscritical data, down from days
Why every pilot gets a baseline and a cost line.
Banking & financial regulation
Systems that had to satisfy auditors
Built a global bank's enterprise metadata repository and end-to-end data lineage for fraud monitoring, supporting regulatory transparency and audit. Earlier, delivered a national financial regulator's market-abuse monitoring platform.
Auditevidence built into the system
Why governance starts in week one, not after launch.
Healthcare · US health insurer
A clinical rules engine for preventive care
Built the engine that matched clinical codes against clinician-written rules to identify members for preventive-care programmes, handling claims data under HIPAA controls. Automated the rule testing that had been the release bottleneck.
1 → 100rule sets validated per month
Why automation earns trust only when experts own the rules.
Work delivered in employed and consulting roles across India, the UK and the US.
How I work
Five commitments you can hold me to.
Pragmatic first
I'm as ready to say "AI is the wrong tool here" as I am to champion it.
Adoption is the deliverable
A tool nobody uses is a failed project, however clever it is.
Honest numbers
I report where AI didn't help too. Credibility is the whole job.
People before tools
Change makes people anxious. I address it openly with your teams.
Independent advice
I don't resell software or take vendor commissions. Recommendations are about fit.
TS
Thanigaivel (Thani) ShanmugamBengaluru
About
Call me Thani.
I'm an AI architect and engineering leader with more than twenty years in enterprise software and data. Most recently, as founding technical leader of an AI platform serving multi-gigawatt renewable energy portfolios, I designed the retrieval, knowledge-graph and multi-agent systems and built the engineering team around them from nothing.
Before that I delivered data platforms and cloud migrations for banks, insurers, healthcare and e-commerce companies in the UK, the US and India. That work taught me how to make technology survive audits, budgets and real users.
Now I help mid-sized companies get the same kind of leadership without hiring for it full-time. I sit with your leadership team, work alongside your people, and build when building is the fastest way forward.
Industries I've worked in
Banking & financial regulation · Insurance · Healthcare · E-commerce · Real estate · Logistics · Renewable energy
Two-minute readiness check
Where is your company on the AI journey?
Six questions. Your answers stay in your browser. You'll see a starting stage and the engagement that usually fits it.
Common questions
What leaders usually ask first.
We're not a technology company. Is this for us?
Yes. Most of the value sits in ordinary work: quotes, orders, inspections, guest messages, reports, invoices. You don't need an AI team to benefit. You need someone who can see those workflows clearly and match them to the right tool.
Do we need clean data before we start?
No. The first four weeks include a data-readiness check. Many high-value uses run on documents, emails and manuals you already have. Where data needs work, the roadmap says exactly what and why.
Will AI replace our people?
The goal is to take repetitive load off your people so they can do more of the work that needs judgment. I work openly with team leads on what changes, train people in their own roles, and build a champions network inside each function.
How is this different from a large consultancy or a software vendor?
You work directly with the person doing the work, from strategy through build and training. I don't resell licences or earn commissions, so tool choices are made on fit and cost. And I stay through adoption, which is where most programmes fail.
What about our customer data, designs and IP?
Governance comes first: a usage policy, approved tools, data classification and review gates, all set up in the first 90 days. Where confidentiality demands it, models can run on your own hardware so data never leaves your premises.
How is it priced?
The Readiness Sprint is a fixed fee. Fractional leadership is a monthly retainer based on days per week. Pilots are priced per pilot. Scope, price and success measures are agreed in writing before we begin.
Where do you work?
I'm based in Bengaluru and work on-site with clients across India. For clients elsewhere I work remotely, with on-site visits for workshops and key milestones.
Start with a conversation
Thirty minutes. No pitch deck.
Tell me about your business and where you think AI might help. You'll leave the call with at least one concrete idea, whether or not we work together.
- 01You describe your business, your teams and what's frustrating today.
- 02I ask about your workflows, systems and data, and share what similar companies have done.
- 03We agree whether a Readiness Sprint makes sense, and if not, what you could try on your own.