THE LAST MILE
Turning AI agents from innovation into impact.
Sushmeet Singh · Founder & CEO, Tesseract Business Solutions
CII AI Powered Summit
Bengaluru · 2026
Who's talking
"I'm not here as a technologist. I ran the business that had to make this work."
The reframe
AI doesn't fail in the lab.
It fails at the handover.
The execution gap
The last mile
90% of the journey is innovation. The impact lives in the last 10%.
Innovation · 90%Impact
The evidence
It isn't just my experience.
It's the pattern.
95%
of enterprise GenAI pilots stall before they ever reach measurable P&L impact.
MIT NANDA · State of AI in Business 2025
40%
of agentic-AI projects will be scrapped by 2027 — on cost, unclear value, and weak risk controls.
Gartner · 2025
And the cause isn't the model. It's the handover.
When agents act
An agent that advises is a tool.
An agent that acts is a colleague.
Advise
Recommends the reorder
→
The question flips
"Is it accurate?"
Do I trust it with authority?
The centrepiece
The Trust–Autonomy Ladder
unlocked by transparency & data lineage
unlocked by explainability & quality checks
unlocked by guardrails, audit trail, role-based authority
unlocked by monitoring, a kill-switch, an accountable owner
↑ more autonomyeach rung earned by governance →
What happens when you skip
Jump straight to rung four,
and the bill arrives.
Cautionary tale · Zillow Offers
An algorithm given authority to buy
Zillow let a pricing model purchase homes at scale — acting, not advising — without the guardrails to catch it when the market turned. It couldn't, and the model kept buying.
$500M+written down
~25%of staff cut
closedthe business
The same use case, governed
Parts reordering, earned rung by rung
Surface the data → recommend the reorder → place it with approval and a full audit trail → only then let it run on its own, watched, with a kill-switch and a named owner.
They didn't lack a model. They lacked the rungs.
The counter-intuitive part
You don't climb by trusting more.
You climb by governing better.
Governance isn't what slows AI down. It's the only thing that lets you speed it up safely.
One conversation
Autonomy and trust aren't two agendas.
They're one conversation.
How you cross it
How you cross the last mile
Strategy
Execution
Operations
Growth
The gap is always here: Execution → Operations.
Proof it works
17%
operational inefficiency removed
15%
OPEX saved over three years
"From instinct to real-time insight."
What it means for this room
Two things to take away
1
If your AI is stuck at pilot, you have a last-mile problem — not a model problem.
2
Last-mile discipline travels. What works on your floor works in a new market.
Innovation earns the headline.
Impact happens in the last mile.
