One in Seven Truck Dispatches with Service Parts into the Field Never Needed to Happen – Interview with Tas Hirani , Global Director Enablement at Aquant I July 2026

One in Seven Truck Dispatches with Service Parts into the Field Never Needed to Happen – Interview with Tas Hirani

What AI actually fixes in the exchange and repair loop, and what it can’t

One in seven truck dispatches with service parts into the field never needed to happen. The part wasn’t broken. The diagnosis was wrong. And someone still paid for the truck, the technician, and the hours.
For six months this series has worked down the service world: availability as a board metric, the org model, the contract, the geography, the data, and last month, where to physically stock parts across a multi-echelon network. This month closes a loop and opens a new one: once the network exists, how do you sustain it through exchange, repair, and pool management, and what is AI actually changing there?
To pressure-test that, I sat down with Tas Hirani, who many of us in the field service industry know personally. Tas has spent more than eleven years in field service and, in her words, “been on more ride-alongs than there are days in a month.” She implements AI for service organizations across very different industries. I come at the same problems from the other side, the logistics and critical service parts supply chain. We agreed to skip what’s on the websites and the marketing and talk about what’s actually hard.

I asked Tas for the costliest myth she keeps having to debunk, and she didn’t reach for a benchmark. She reached for a story.

A large fulfillment operation prints books on demand. Two industrial printers sit side by side, running around the clock. When one goes down, a rotation of engineers is on site 24/7 to get it back up. The fix they reach for is a print head that costs $150,000. The engineer on the next shift hits a problem and replaces the print head again. The shift after that, again.

“Within 24 hours you actually replaced a huge value in parts. You’re getting close to half a million dollars, which is insane when you think about it, but it’s genuinely what happens in the real world.”

The industry has an ugly name for this: parts shotgunning. Swapping the expensive thing is easier than getting to the root cause of why it failed. And, Tas is careful to point out, no individual technician is doing anything wrong. Each one did the job as prescribed. What was missing was the layer above them: data capture and communication across shifts. Three people solving the same problem because none of them could see that the previous two already tried.

That is the thesis of this entire installment. The exchange and repair loop does not break on technology. It breaks on data and communication. And that’s the point: it is also the reason AI both helps more than the skeptics expect and less than the brochures promise.

No-Fault-Found Is a Data Failure, Not a Parts Failure

The single biggest cost driver in any exchange pool is the part that comes back with nothing wrong with it: no-fault-found. It is the most studied number in service and still the most stubborn. Tas reframes it in a way that makes us think:

“When there isn’t a fault found, it’s the symptoms that led to that conclusion which don’t often get recorded. That then means that you continue to have that problem.”

No-fault-found isn’t an event. It’s a recurrence machine. You pull the part, find nothing, return it to the pool, and because the symptoms that triggered the swap never made it into a system anyone can query, the same symptom sends the same part back next month. The pool balloons. The math falls apart. Everyone assumes someone else is managing the loop.

Interestingly, Tas insists the cost isn’t only financial:

“The engineer that is actually out and about on his day, he loves problem solving. There’s no problem to solve in this situation. It’s very frustrating. So not only is it financially frustrating, it’s also from a human side, frustrating too.”

If you want to know whether a service organization’s loop is healthy, Tas says, watch for one sign: urgency. The broken ones feel “very chilled, very relaxed, very calm. It doesn’t really matter if it takes a few extra days.” Sometimes that calm is earned, because a swap unit covers the customer. Often it isn’t.

From the logistics side I see the same failure from the opposite end. We will run a part across a region in two or three hours because everyone is ringing the bell, and there is nobody to receive it, or no technician has been assigned to the case. The urgency was real on our end and absent on theirs. The loop leaks at both ends, and the customer pays for the gap in the middle.

What AI actually changes: visibility, not magic

So an organization with a broken loop adopts AI. Walk through what changes. Tas’s answer is deflating in the best way. It isn’t a model. It’s a parcel-tracking screen.

“We’ve all got really used to understanding where our parcel is and how many stops away it is… if you think about the kid in a stroller with the iPad, if they can understand where an item is, then anyone can.”

The win is visibility, surfaced simply enough that the whole chain can coordinate around it: the technician who needs to be there, the person who needs to take receipt of a critical part so the repair lands inside a tight SLA window.

The simplicity is one of the most needed features in any market today. That simplicity does something underrated in field service. It bridges a workforce that spans generations. The most common way good data dies in the field is a 60-year-old master technician staring at a tiny screen and a drop-down menu, giving up, and leaving the failure note blank. Voice and plain-language interfaces don’t just speed data entry. They are the difference between the loop being fed and the loop failing.

But visibility assumes data, and in exchange and repair the data is famously messy: incomplete failure codes, inconsistent technician notes, missing asset history. What happens when you train AI on that? Here Tas draws the line that separates the organizations that get real dollar savings from the ones that buy software and stall:

“The companies where I’ve seen the best adoption are the ones that have actually been thinking about the content needing to be AI-ready for a very long time… Think about it like foundations for your house. If the foundations are temporary, you’re going to get a house that’s only going to work for a short period of time.”

Crucially, “AI-ready” is not a reason to wait. You can use AI itself to plug the gaps, she argues, if you’re smart about it: detect that an incoming fault code is wrong, restructure documentation into a form the retrieval layer can actually use, and improve the data layer as you go. AI-readiness is a starting move, not a prerequisite.

The half of the loop AI doesn’t touch

This is where we supply chain professionals have a lot to say, because it’s the half we logistics service heroes live in. AI improves diagnosis and triage. It does not move a part. The right unit still has to arrive at the right place, in the right condition, to the right technician, through the right network, inside the SLA. That is physics and logistics, not technology.

Tas’s answer was honest: the proof is in the data, the companies that adopt correctly get results, and the worst thing you can do is try to do it in isolation. That is true. And it also confirms the division of labor this series keeps returning to. AI is rewiring the decision layer of the loop. The physical layer is still a supply chain problem. A perfect diagnosis does not help if the part cannot physically reach the technician in time, and that is exactly where the loop tends to break.

The client stories we meet too often in the beginning of onboarding: a customer sells equipment into a data center, the data center later moves 100 kilometers north, and nobody updates the install-base record because nobody told anyone. Then a 4-hour SLA fires, the corrected address surfaces during the call, and no AI on earth lets you pick up and prepare the service parts and drive them 300 kilometers in four hours. The decision was right. The loop still failed, because the physical network was designed against a fact that was no longer true. Predictive triage is worthless if the address, the stocking point, and the route can’t deliver against it. The decision layer and the physical layer have to be designed backward from the same outcome, or you get a smarter system that fails in the same place.

Why many AI implementation projects die, and who actually has to say yes

For organizations that fail, Tas says the cause is rarely the technology. It’s a finance story told wrong.

“There is this misconception that you’re going to reduce headcount straight away because you’re introducing an AI solution. And that is not true at all. If that’s what you’re looking for, to hit the P&L straight away, I think that’s setting you up to fail.”

The real return shows up slowly: a lower need to backfill as people leave, faster onboarding, fewer recruits required to grow. Leaders who can tell the short-term-versus-long-term story honestly keep their budgets. Leaders who promise an instant headcount cut lose theirs the first quarter it doesn’t materialize.

And the hardest person to convince? Her answer was immediate: “It’s normally IT.” Where she’s seen real success, the person asking for the technology has a direct line to the top of the IT chain. “If that relationship does not exist, they’re never going to get an AI solution.” The C-suite then divides cleanly:

• CFO wants clarity. Bring a few key metrics, not six or eight, or finance will hold you to all of them.
• COO wants operational efficiency: no-fault-found down, first-time-fix up, average repair time down, especially on multi-day SLAs with parts shipments in the critical path.
• HR / People Officer wants retention and attraction. An AI stack is now a recruiting argument: candidates ask what tools and what onboarding they’ll get.

And no, you don’t have to be big

Most benchmark data comes from large organizations, so the obvious worry for a mid-size operator, under 100 technicians, is that this is a game for the giants. Tas flipped it: “They need to be smarter in order to grow.” Her smaller customers are already running vision and voice. A technician completes the work order by voice from the van, hears the asset’s full timeline before arriving, opens the machine, gets a video feed that identifies the part number visually, and authorizes the order on site. Here we both agree, for the small and mid-size operators we serve, the ones trying to play in the logistics league of the big guys, that is the leverage that closes the gap.

A closed-loop health check

A loop you can’t measure is a loop you can’t defend. Use this before your next exchange / repair review.

Triage gate (before a part is touched):

1. Are the symptoms that triggered this swap recorded in a system someone else can query? (If no, you are manufacturing your next no-fault-found.)
2. Has this asset’s full history been read before work starts, not after?
3. Can every node in the chain see where the part is right now, in plain language?
4. Is anyone confirmed to receive the part and is a technician assigned, before it ships?
Pool-health metrics to track (manage the tail, not just the average):

Pool-health metrics to track (manage the tail, not just the average):

No-fault-found rate

Healthy loop: Falling quarter over quarter

What it exposes: Diagnosis and data capture quality

First-time-fix rate

Healthy loop: Rising

What it exposes: Whether the right part or skill reached the site

Average repair / turnaround time

Healthy loop: Falling, especially on multi-day SLAs

What it exposes: Physical loop speed

Pool size vs. installed base

Healthy loop: Stable or shrinking per unit

What it exposes: Whether shotgunning is inflating the pool

Symptom-capture rate on returns

Healthy loop: Approaching 100%

What it exposes: The root cause of recurring no-fault-found cases

Install-base address accuracy

Healthy loop: Audited, not assumed

What it exposes: Whether your SLA is physically deliverable

AI-readiness check:

• Is your content structured for retrieval?
• Is there a direct line to the top of IT?
• Have you prepared your people, not just your data?

Two actions for next quarter

1. Spend the quarter on the problem, not the tool. Tas’s parting advice was the sharpest line of the interview: “For the 90 days, spend 89 days thinking about what your problem is and only spend one day thinking about solutions. People jump to solutions far too fast… when you try and fit the solution around your perception of the problem, that’s where things go wrong.” Before you scope a single vendor, instrument your loop with the metrics above and find where it actually leaks.
2. Close the symptom-capture gap. Pick the one return reason that recurs most and make recording its symptoms mandatory and effortless (voice, not drop-downs). It is the cheapest move with the largest effect on no-fault-found, and it is the data foundation everything predictive will later stand on.

There is one frontier Tas flagged that most teams aren’t ready to look at yet: conversational AI in the loop itself, phone calls and interactions handled by an AI chatbots that “drives an intensity in the speed of delivery that people just underestimate.” It makes people uncomfortable, some fear it may replace them. But it is also the next real game changer.

Closure

For a decade we treated exchange and repair as the unglamorous back end of service. It is quietly becoming the place where AI and supply chain either meet or miss each other. The organizations that close that gap first will not just spend less. They will be the only ones who can actually promise what their contracts already claim.

Guest: Tas Hirani, Global Director Enablement at Aquant

Author: Eyal Yossef, VP Supply Chain Solutions at Unilog