Leadership Spotlight

Decision Intelligence is The Next AI Wave with Sandeep Khemami

AI enables better judgment. And gives judgment better evidence, faster.

— Sandeep Khemani

A Conversation with Sandeep Khemani, Senior Vice President and Corporate Studio Lead

For years, organizations have been collecting more data than ever before. Yet for many leaders, more information has not necessarily translated into better decisions.
Sandeep's team builds solutions the market actually wants and works with clients to define the use cases of the future, not the past.  Rather than focusing solely on automation, he believes the next wave of AI adoption will be driven by something far more valuable: decision intelligence.

For years we've been talking about "big data." Suddenly everyone's talking about decision intelligence. What changed?

Two things. First, AI can finally read what was hard to interpret or digest: contracts, PDFs, images, videos and recordings. Second, we can give AI a map of the business, built from the client’s own definitions of the business processes. For example, suppliers, parts and products (an ontology) and relationships between them. That is a knowledge graph. Agents follow the connections and reason with facts instead of guessing, while large language models do the reading and explaining.

That moves AI from insight to AI-enabled decision intelligence: not just what happened, but what it means and what your options are, grounded in your data. With predictive intelligence, machine learning that shows what could happen next, it also looks ahead, so you can reduce risk and take ethically available opportunities.

Where does that show up in business?

A late supplier, seen early is a good example.

Machine learning flags that a supplier will likely ship three weeks late. The AI follows the map: two parts, two products, 14 customer orders at risk. It also finds a second approved supplier for the same parts at a higher price, and lays out the options in minutes, before the delay hits. A person decides.

The same pattern works across any industry domain, be it supply chain, healthcare, manufacturing, banking or retail. Our solutions aim to grow our clients’ top line, or improve productivity and cut costs. For example,

  • Supply chain and procurement: cleaner data, find savings in the network, invoice processing accuracy.
  • Product safety: ingredient evidence organized, so safe products launch sooner.
  • Getting drugs to market: evidence assembled for regulators and payors, helping drugs reach patients faster.
  • Intellectual property: patents watched, whitespace spotted: protect what you own, find where to innovate.

AI can also surface future revenue streams that were not visible before. Innovation is the name of the game.

There is a lot of discussion about AI replacing work. Do you see it differently?

I do. I think of agents in three tiers, and agentic autonomy has to be earned:

  1. Act alone, repetitive, low-risk work that is manual and boring, but important, like invoice reconciliation.
  2. Recommend, person approves, anything with a financial or other material impact, or that is hard to rollback, such as a payment, or new supplier terms.
  3. Learn from approvals, handles routine cases people always approve, within limits, and escalates the rest, routine exceptions.

AI is excellent at repetitive tasks and repetitive judgment. People stay essential for context, exceptions, priorities and accountability. That is where this is going: agents earn more autonomy, one decision at a time. 

What is the core of Domain-Led AI? Why are domain experts becoming more important as technology progresses?

Because AI doesn’t automatically know how your business works.

Every organization has specific rules that differ, even within competitors in the same industry. And a pharma company does not operate like a manufacturer, and a process in Europe can differ from the same one in the United States. Domain experts define those rules, and agents operate within those guardrails.

Here is one rule we hold firmly: an agent must never make an autonomous decision with a business impact that cannot be reversed. A person stays involved, at least until confidence in the agent is extremely high. Our AI governance and security posture is extremely tight in the same spirit: client data stays the client’s data, AI only sees what the person asking is allowed to see, and every answer shows its sources and every action is recorded.

What's one AI metric executives spend too much time looking at?

Token consumption. It is easy to measure and without effective solution design, can leave costs exactly where they were. With the right AI solution design, the right metrics are business outcomes you set out to improve – for example, faster processing, lower inventory, better accuracy, higher sales.

That is why value realization is part of how we design and deliver our industry solution. Evalueserve’s AI Value Realization framework starts by identifying today’s metrics, then asks three things in order: will people operate more productively, can everyday decisions be better, and is the business better overall? For example, a finance team starts with 18% of invoices needing manual work and a target to reduce that to 8%. Each month there should be tangible progress towards that target or remediation be implemented.

Looking ahead, what do you think leaders are underestimating today?

The exceptions, and the people.

Most AI looks great on the clean path: the complete document, the matching number, the rule that applies. Value is won or lost in the exceptions, such as the missing document, the conflicting data or the unusual customer. Tools are easy to get, but people need to understand what AI is good at, where it needs oversight, and how to challenge its answers – all to make the solution better. 

What's one belief about AI that you think executives need to challenge right now?

That deploying AI is the hard part. For application development, technology is usually the easiest piece. The hard parts are knowing where AI fits in a process, setting up governance and controls, preparing people, and deciding how success will be measured.

That is why we begin by asking whether value will be realized for this client. The differentiator will not be who adopts AI first. It will be who integrates it responsibly into the way work gets done. 

As AI becomes embedded across business processes, Sandeep believes success does not relate to who has the most AI tools. The successful approach is to combine technology, human expertise, and strong governance to make better decisions every day. AI’s greatest contribution may not be to do the work for us. It will be help us do it better.

Latest Posts

Talk to One of Our Experts

Get in touch today to find out about how Evalueserve can help you improve your processes, making you better, faster and more efficient.