The Trust Ladder: Amit Mehta on How Enterprises Learn to Let AI Act on Their Behalf

In this conversation with Amit, he shares with decision-makers the real reason behind pushback on enterprise adoption.

How Has AI Changed AQe Digital and the Businesses it Serves?

I started AQe Digital in the age of COBOL, and now we are in the age of AI. I have watched several technology waves arrive with a lot of noise. AI feels different because it does not stay inside one department. It reaches into every kind of work we do.

Today we are a team of more than 650 people using AI across multiple business verticals like software consulting, product engineering, AI and ML solutions, end-to-end digital services, AEC and building services, and publishing services. Our clients are not limited to one region scattered across the world and are from different industries. 

What this meant was we wanted AI adoption at every stage, every deliverable, and most importantly, at every process. One team was bringing AI, IoT, and automation onto the shop floor. Another team adopted AI, automating data conversion and tagging for the publishing companies. Plus, our teams were using data to understand buyers better, enabling AI adoption for retail companies.  Each department has made its own approach and framework to get the expected result.

The bigger lesson was about pace. We learned to put AI to work in our own delivery first, see where it held up and where it needed a person watching, and only then take it to clients. That is the same discipline we now bring to businesses in manufacturing, healthcare, retail, publishing, energy, and real estate. It is also the thinking behind everything I will share in this conversation.

Why do Enterprises Doubt AI Adoption?

Using AI and delegating to AI are two different decisions, and most of us have only made the first one. Letting a model summarize a contract costs you nothing if it gets it wrong. Letting it approve a refund, reorder inventory, or reply to a regulator does. 

That second decision touches accountability, and accountability sits with us. For enterprises moving from AI experimentation to production-ready AI, the challenge is not simply choosing a model; it is building the data, governance, and operational foundation that allows AI to work safely at scale.

Removing the human entirely from the decision-making process and providing autonomy to AI is where many organizations draw the line. And  I have sat in boardrooms where the AI roadmap listed several use cases and not one line about who answers when an agent makes a bad call. That is the real blocker. Not the technology, but the absence of a design for trust.

Where did the Idea of a Trust Ladder Come From?

Over the span of 29+ years, I have noticed something peculiar about processes and people. Autonomy and direction help shape the right processes, and with the right process, your teams deliver results you envisioned. This observation made me curious. Can I apply the same logic to AI adoption?
What if we design processes that allow the AI systems to function with autonomy and the right direction, delivering results that align with organizational goals? And this led to another question: what framework can we build for teams to trust AI and deliver what’s best for our clients? 

The answer was, “Trust Ladder”

The ladder bridges the gap between AI as a tool that drafts emails and Agentic systems. With the trust ladder, a leadership team can see where each workflow sits today and what it would take to move it up. But a trust ladder is incomplete without its rungs. 

What are Rungs: Can You Please Walk Us Through?

There are five, and I would encourage every business leader  to map their own AI portfolio against them.

1. Inform. AI summarizes, flags, and answers. People decide everything.
2. I recommend it. AI drafts the action. A person approves or edits it.
3. Act with approval. AI executes, but only after a named person signs off.
4. Act within guardrails. AI acts alone inside a hard boundary, such as a spend limit, a customer tier, or a region. People audit samples.
5. Act and report. AI runs the workflow end to end. People handle the exceptions.

Most enterprises I speak with have dozens of use cases on rungs one and two, a handful on rung three, and almost nothing above that. It is not a lack of ambition. It is that nobody has defined what climbing actually requires.

What Has to be True Before a Workflow Earns the Next Rung?

I ask four questions, and they are the same ones I would ask before promoting a person.

• Track record: Has the agent been right, consistently, at its current rung, and do we have data to prove it rather than a feeling?
• Reversibility: If it gets this wrong, can we undo it in minutes, or is the damage permanent?
• Blast radius: How many customers, how much money, or how many compliance obligations does one bad decision touch?
• Visibility: Can someone see exactly what the agent did and why, after the fact?

If any answer is weak, the workflow stays where it is not forever, just until the team fixes that answer.

This turns into “do we trust AI?” from a boardroom debate into a backlog your teams can actually work through.

Why is Trust Granted per Workflow, Not Per Company?

Because the same company can be ready for rung five in one place and rung one in another, and that is healthy. An agent that reconciles low-value invoices can run largely on its own. An agent that touches patient records or credit decisions should climb far more slowly, even inside the same organization, even on the same model.

The mistake I see is a company-wide AI policy that says “agents may do X.” It ends up too loose for your riskiest workflow and too tight for your safest one. Set the rules at the workflow level, and let each workflow earn its own position on the ladder.

Which Mistake is More Common: Climbing too Fast or Never Climbing?

Both happen, but never climbing is quieter, so it does more damage. A team that jumps to rung four too early has an incident, and everyone learns from it. A team that parks everything on rung two never has an incident, and never sees a return either. Twelve months later, the CFO asks where the value went.

Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027, citing rising costs, unclear business value, and inadequate risk controls. I read that as two failures at once. Some projects climbed without guardrails. Others never climbed high enough to pay for themselves.

Who Should Own the Decision to Give an Agent More Autonomy?

The business owner of the workflow. Not IT, and not the AI team. IT builds the guardrails and risk defines the boundaries, but the person who would answer for the outcome if a human made the same call should approve the promotion.

In practice, I recommend a short promotion review, much like a quarterly performance review. The workflow owner, someone from risk, and the technology lead look at the evidence and decide one of three things: hold, climb, or step back. Stepping back has to be a normal outcome, not an embarrassment. If you cannot demote an agent easily, you will always be afraid to promote one.

How Does AQe Digital Help Enterprises Adopt AI with Confidence?

AI adoption shouldn’t require enterprises to choose between moving fast and staying in control. AQe Digital helps leadership teams bridge that gap by building the governance, safeguards, and operational foundation needed to scale AI with confidence.

Our role goes beyond building AI agents or individual use cases. We help enterprises create the scaffolding around AI that makes adoption secure, measurable, and accountable, from decision logs and rollback mechanisms to approval checkpoints and clearly defined boundaries. Our AI/ML development services help turn these requirements into production-ready systems.

AQe Digital also helps enterprises map their AI landscape, evaluate use cases by business impact and risk, and identify the workflows where AI can deliver the greatest value. This AI strategy and consulting approach helps organizations move from experimentation to a roadmap designed for scale.

By prioritizing the right use cases and defining how each one should progress from experimentation to production, organizations can avoid disconnected pilots and build an AI roadmap designed for scale.

This approach allows enterprises to move from AI experimentation to trusted, business-ready adoption with AQe Digital as the technology partner that helps them take every step with greater confidence.

What Should Business Leaders Do in the Next 90 Days?

Three things.

1. Map it. Ask your team to place every live AI use case on one of the five rungs. You will likely find most of them on the bottom two.
2. Pick one workflow. Choose one where errors are cheap and reversible, but volume is high, and set a target to move it up one rung this quarter.
3. Write the demotion rule first. Agree in writing what would make you step the agent back down. Teams move faster when they know the exit exists.

What Would You Tell Fellow Business Leaders Who Are Still Skeptical?

Keep your skepticism. It is an asset. Just point it at the right question. Do not ask whether AI can be trusted. Ask which decisions, within what limits, with what proof. You already know how to answer that, because you answer it every time you delegate to a person.

None of us built our companies by doing everything ourselves. We built them by learning whom to trust with what, and by checking. AI deserves the same discipline. No more, and no less.