The clue was a milk crate. At college, Stephen Carrabba watched one person hunch over a microscope while another stood on the crate to reach it. The instrument cost about $15,000. The bench beneath it was fixed, ordinary, and indifferent to the humans asked to use it. Most people would have adjusted their posture and continued with class. Carrabba adjusted the premise.
That small indignity became the AdjusTable, a height-adjustable laboratory workstation developed through his first company, Accuposture Systems. It also became a neat preview of the next two decades. Carrabba has repeatedly wandered into systems that appear settled, noticed an awkward workaround, and asked why everyone has agreed to live with it.
Today the thing under inspection is no longer a laboratory bench. It is the machinery that pays healthcare claims. Carrabba is the co-founder and CEO of ClaimInformatics, a Bloomfield, Connecticut company that reviews claims before and after payment for self-funded employers and the firms that administer their plans. The language has changed from pivot points and stainless-steel hardware to claim lines, coding edits, contracts, and fiduciary process. His favored move has not: open the casing and look inside.
The useful discomfort
Carrabba studied entrepreneurship at Babson College, graduating in 2005. In a 2009 career interview, he described a childhood habit of taking apart gifts to see how they worked, then wondering how their design might improve. It sounds charming until you imagine being the relative who bought the toy. It also explains the direct route from an uncomfortable lab class to a manufacturing business.
The first workstation was no napkin sketch carried triumphantly to market. It took sixteen prototypes. Fourteen pivot points required a precise, economical bearing, and no standard part did the job. Carrabba's team engineered an acetal bearing of its own while also working through requirements involving sanitation, accessibility, electrical safety, and furniture standards. The eventual workstation received a best new ergonomic product award in 2007.
There was a more painful education. Eager to launch an early design, Carrabba put units into production without fully prototyping them. Dissatisfied with the result, he scrapped the run. The penalty arrived in the traditional entrepreneurial currencies: time and cash. His correction was admirably unromantic. From then on, prototype fully and test every function.
“Problems create possibilities.”Stephen Carrabba, 2009
It is tempting to polish this into a parable about glorious failure. Carrabba's own recollection was less decorative. Failure was tough. Each rejected prototype moved the product closer, but each one also had a cost. The practical lesson mattered more than the mythology. Curiosity starts the work; disciplined testing keeps curiosity from becoming an invoice.
From physical joints to expense lines
The next business exchanged fabricated parts for contracts and invoices. Carrabba founded Expense Consulting, a performance-based cost-reduction consultancy. Its work eventually concentrated on post-acute and acute-care organizations, but that focus arrived organically. The firm began by helping larger private practices in Connecticut and New England. An accountant introduced the team to a skilled nursing facility in 2015, and a niche took shape.
The raw material was gloriously unglamorous: waste streams, laundry, food service, utilities, laboratory costs, radiology, pharmacy, and cloud technology. Carrabba described starting with a historical look at where dollars went, then examining service levels, pickup schedules, tariffs, and contract terms. This is entrepreneurship conducted with a flashlight in the utility cupboard.
One client supplied the perfect explanation for the model. How many waste-disposal contracts had Carrabba's team seen in three years? Dozens, perhaps hundreds. The client had seen one. Repetition created expertise, and expertise made the overlooked legible. Carrabba said 82 percent of the consultancy's projects stayed with incumbent vendors. Savings did not automatically require a dramatic switch; sometimes the useful act was simply knowing what the existing relationship ought to cost.
One operating instinct, three applications
Carrabba's public comments from this period reveal an operator who likes alignment more than theater. He emphasized autonomy, professional communication, and recommendations that accounted for the cost of changing vendors. “We never compromise quality; it is not all about the dollars,” he said in 2024. The line matters because cost reduction has a natural temptation to celebrate subtraction. Carrabba framed the better result as precision: neither overserviced nor underserviced, neither loyal from inertia nor disloyal for show.
A claim is a tiny contract drama
The path to ClaimInformatics began inside Expense Consulting. Carrabba's team built technology to audit pharmaceutical claims for hospitals and nursing homes. Through that work he met Dawn Cornelis, who would become a co-founder. The proposition widened: if software could review pharmacy transactions, why not examine medical claims for improper payments? Carrabba liked the idea, built a business model around it, invested, and began operating the company in 2017.
The first work was retrospective. ClaimInformatics looked back at payments already made, searching for errors and recoverable money. It later added pre-payment review, checking transactions before plan funds leave, and ongoing post-payment monitoring. In Carrabba's telling, the two modes belong to one continuum. Finding yesterday's error can recover money. Catching tomorrow's version can prevent the chase.
This work rewards the same temperament as product prototyping, only the exploded diagram is now made of billing codes, plan documents, network arrangements, and administrative agreements. A dashboard can say what was paid. It cannot, by that fact alone, establish that the amount followed every relevant rule. Summary is useful. Verification lives further down.
What sits beneath the summary
Company-reported scope for the ClaimIntelligence edit suite. Bar lengths are illustrative; the labels show the reported figures.
Carrabba's public argument increasingly centers on independence. In the market he describes, carriers, third-party administrators, brokers, networks, and recovery vendors may have incentives that do not perfectly match those of the employer funding a plan. His remedy is not eloquent trust. It is access to the data, a review process, and records showing what was checked.
That stance also shapes the company's pricing pitch. ClaimInformatics has promoted a flat, per-claim fee for pre-payment review rather than taking a percentage of identified savings. Carrabba argues that a vendor paid more whenever errors recur has an awkward incentive to preserve the source of its revenue. A fixed transaction fee is his proposed answer: pay for the checking, not for the size of the mistake.
“Healthcare claims are one of the biggest blind spots in employer benefits.”Stephen Carrabba, 2026
Curiosity with guardrails
In 2026, Carrabba brought the same skepticism to artificial intelligence. On the Broken Healthcare podcast, he described testing a long contract prompt across generic models and getting inconsistent answers. The company then spent eight months building a constrained system around a large collection of legal material and a fixed rubric of questions. His verdict on unstructured inputs was pleasingly blunt: “Garbage in, garbage out.”
The episode offers a useful counterweight to fashionable AI claims. Carrabba is clearly interested in early adoption; his industry biography emphasizes his habit of researching new technology. Yet his old manufacturing lesson survives inside the newer software story. A powerful tool is not excused from prototyping. Repeatability matters. The rules matter. The system must be shaped around the job, much as a custom bearing once had to be shaped around fourteen stubborn pivot points.
By 2026, his role had expanded beyond building the product. He spoke at the SIIA Healthcare Price Transparency Forum on AI, APIs, and usable claims data. He joined a PLANSPONSOR livestream about data-driven payment-integrity decisions. ClaimInformatics added a chief operating officer and continued hiring around partnerships and operations. Carrabba remained the hands-on founder described in conference biographies: involved in technology, talent, strategic relationships, and the direction of the business.
The Carrabba operating note
Look for the workaround people have stopped seeing. Trace it to the underlying mechanism. Build the simplest useful correction. Test it before scale. Then check whether the person paid to solve the problem benefits from letting it continue.
The question beneath the question
Carrabba's career is easy to divide into chapters: ergonomic products, expense consulting, claims technology. It is more revealing to read it as one long act of inspection. The lab student should not need a milk crate. The client should not need to become an expert in a contract encountered once. The plan sponsor should not confuse a completed payment with a verified one.
There is also a pleasingly physical quality to his way of thinking. He has listed metal fabrication and woodworking among his hobbies. Both punish vagueness. A joint fits or it does not. A surface is level or it is not. In the world of healthcare administration, certainty is harder to touch, but Carrabba keeps reaching for the equivalent: the applicable rule, the claim line, the contract clause, the documented review.
The ambition now is larger than a better bench and less visible than a piece of furniture. Carrabba wants claims to be transparent enough that fiduciaries can show their work. It is not the sort of mission that produces a glamorous object. It produces a better question, asked before the money moves.