Three Conversations, One Missing Decision

Signal Labs CEO Rajeev Ronanki recently had three different interviews, sharing thoughts on hiring ledgers and customer service automation, as well as pilot zombie land. All rooted in company actions based on what AI takes away.

"How does $25 million buy BrassRing?"

That was Chad Sowash's tongue-in-cheek comment on the recent acquisition of BrassRing, the ATS and talent acquisition platform that runs recruiting for many enterprises, including plenty of Fortune 500 organizations. It came nine minutes into a high-energy Chad & Cheese interview with Signal Labs founder and CEO Rajeev Ronanki.

In the video below, Raj corrected the premise, and the correction turns out to be the better story.

The turnstile: what the $25 million actually was

The $25M wasn’t the price of the acquisition, but early total contract value already attached to Signal Labs, when it launched in April 2026. Backing comes from Lightspeed Venture Partners as well as private investors, and a handful of marquee customers signed on as design partners when Signal Labs came out of stealth. As Raj described it, partners such as Infinite Computer Systems "made a bet on Signal Labs," on the view that an asset they had treated as one worry among many could be front and center somewhere else.

What the deal actually buys is a turnstile. For more than a quarter century ATS and talent platform leader BrassRing processed more than 350 million applicants, recording two things that matter most about a workforce over time: who came in, and how long they stayed.

This turnstile becomes useful the moment enterprises stop focusing on headcount and start accounting for capacity. The idea traces back to a customer conversation, where Signal Labs clients kept naming BrassRing as the system they used for hiring. Raj asked them what happens when they put AI agents next to people, with physical AI & robotics arriving later, and was all of it being managed on a single ledger.

The answer, from many different HR leaders, came back as crickets.

Nobody should read that as HR failing to keep up. It matches Raj's description of the function being handed "an order to go do something" instead of shaping the workforce as a thought partner. A turnstile only earns its keep if somebody in the building has decided what the building is for, which is the subject of the second conversation.

Pilot zombie land: why enterprise AI pilots stall

This clip tells a story, crucial for company leaders. Today’s enterprises are burning through their annual AI token buget in just a few months. One well-known enteprise spent $500 million in just 30 days. This is because many organizations have no AI spend caps; its surprising, as roughly 90% of companies are not seeing ROI on their spend.

Only 22% of 1,303 companies surveyed have scaled AI across multiple business units, Gartner reported on September 1, while 85% of functional leaders plan to spend more in 2026. An agentic reasoning model also costs at least five times a basic chatbot per task, with cost per workflow rising fivefold through 2028.

Raj sees two related problems:

1. You take on an AI project and cannot find the ROI in it, or
2. you get stuck in what he calls pilot zombie land, running lots of interesting proof of concepts and pilots that never scale: a hangar full of aircraft, fueled and staffed, none of them cleared for takeoff.

Both come from "trying to shoehorn AI into the existing structures of an enterprise." These are structures that are older than many of the people who work in them. Departments, processes and functions date back to the 1980s, when computers arrived to automate paper-based workflows.

Raj reaches for Mad Men, making it physical: work shuffled between floors in big brown envelopes until computers made that shuffling obsolete. The envelopes are gone, and still the floors they traveled between remain standing.

When AI takes over more of the doing, the coordination or ‘talking’ between departments drops off, and the context goes with it. Every data silo gets deeper, with better tools and worse hearing.

"So inadvertently AI is leading to a level of incoherence because of the way that companies are organized. The work itself is organized incorrectly to put AI into it."

Raj Ronanki, CEO, Signal Labs

Emily Binder, Signal Labs CMO, and the inteview host, recalls the early 2000s rush to mobile optimize websites. Taping a desktop layout onto a phone screen gave companies a meaningful lesson that the interaction itself had changed.

Raj refuses the framing that treats AI as the newest tech upgrade. Instead, he takes a bold step further, saying "It's fundamentally a new form of intelligence," and giving each department AI with the wrong framing wastes it.

Seventy percent of the calls, and the wrong target

The third conversation shows what happens after a company does automation well. Here, Raj is with Milind Shah, Head of Payer at ServiceNow, who worked with him at Elevance.

At the large payer, in 2021 and 2022, they set a goal of automating roughly 70% of inbound calls. This was before generative AI, and they pulled it off. Today, Raj believes the more relevant question is, “What level of automation should be the target.”

For a health plan, the contact center is close to the primary signal on how members think and feel about their care. Raj wants it reading leading indicators on Stars and quality, which makes the deflection target look small: "we get twenty million calls a year, how do we get rid of ninety percent of them... that's somewhat of a short-sighted goal."

What he proposes is human plus AI serving the customer together, and the headcount conclusion runs the other way from the usual one. "Perhaps you need to expand the number of agents versus reducing that."

The through line

Three conversations, one habit: measuring AI by what it takes away.

On the recruiting side companies count headcount, then pay to bring more than half of those cut roles back six months later. Those numbers are made clear in our recent post ‘AI Rehiring Whipsaw.’ Pilots get counted by the dozen while only 22% of enterprises have scaled AI across business units. In the contact center of a large payer, deflection hit 70%, and the reality that twenty million calls give a health plan on member sentiment never entered the scorecard.

Raj names the cause himself in the second conversation, and it is structural. "You're not organized in a way to take advantage of the technology just yet," he says, which makes redefining the operating model what he calls the missing link. An enterprise organized to move paper can automate paper faster, and it still cannot say what the work is for.

Signal Labs built a category of software to answer that question: Systems of Attention.

Which brings the argument back to the twenty million calls. Deflecting ninety percent of them dismantles the plan's listening post to save on the electricity. Plenty of enterprises are signing off on that trade this quarter, and very few have written down what they agreed to stop knowing.