HKR.TEAM: Do the job. Then redesign it.

Jamie has a job in finance and a rather useful disability: it cannot mark an invoice paid. At HKR.TEAM, the AI agent checks for incoming money, proposes matches and hands the consequential click to a person. In the company’s account, the arrangement replaced checking bank apps whenever somebody found a moment. The interesting detail is the missing permission. A business selling automation has drawn a line around what its automation may do.

That line helps explain HKR.TEAM’s latest direction. After years building remote teams, it now offers a “Forward Deployed Operator”: someone who does a job, learns its peculiarities and stays to run the AI built around it. The proposition puts practical knowledge ahead of software. First, understand why the invoice looks wrong. Then decide which part of checking it belongs to a machine.

01 The first product was a person

The company’s published origin begins in 2016 with Sergiu Poenaru working remotely for a US business. Asked to find more people who worked the same way, he began assembling a team in Romania. Referrals brought more founders; customer service widened into operations and sales. There was a simple source of demand: a client who trusted one assignment had another assignment waiting.

Sergiu is CEO; co-founder Andreea Poenaru is chief of staff. The current About page describes HKR as bootstrapped, profitable and founder-led. Its culture statement mixes serious work with an affection for internet nerds. For a company whose people spend their days inside other companies, that combination feels apt: be comfortable online, and take the assignment seriously.

02 One hire earns the next

TeePublic supplies the long version of the experiment. HKR dates the relationship to a single affiliate-marketing hire in 2016. It later expanded into customer support, moderation and account work. By the holiday season recalled in its customer interview, the team had reached 122 people. The company reports roughly $2 million in annual payroll savings against equivalent US hiring, alongside 97% customer satisfaction.

Those are results from HKR’s case study, rather than a forecast for the next buyer. What travels better is the sequence: begin with useful work, establish trust, expand when demand arrives. An anonymous luxury design client tested two sales roles for 90 days. HKR reports more than 200 qualified leads from over 2,300 accounts contacted; the client then increased the team to five. Research and outreach had separate owners, with a documented HubSpot pipeline connecting them.

03 Twelve seats do not guarantee twelve good answers

At Sakara Life, the problem was more personal. The premium nutrition brand had roughly 12 support agents across an internal team and a traditional provider. Replies leaned on macros. Some customers thought they sounded robotic, although no AI was involved. The mismatch was between what the brand sold and how it answered people.

HKR describes a replacement team of five, recruited with Sakara’s direct involvement and supported by ongoing coaching. It reports about a 10% increase in customer satisfaction and a first response almost 90 minutes faster. Customers began requesting particular agents. A smaller team had become more recognizable.

The lesson is fairly demanding. Hiring fewer people only helps if recruitment, coverage, coaching and authority improve together. Shrinking the roster by itself does not reproduce Sakara’s experience.

04 Automation can make its own mess

HKR’s internal sales experiment offers a useful complication. A meeting-notes tool had created tasks automatically in its CRM. They accumulated beside the real work. In August 2026, HKR says its newer sales agent, Ari, cleared 289 old tasks at Cristina Bufnea’s request; about 90% had come from that earlier tool.

Ari now drafts deal notes, stages and next actions from conversations in Slack. A person instructs or approves the update. Cristina, head of partnerships and accounts, keeps client decisions and sends follow-up emails herself. The lesson is about cleanup as much as speed: a system that creates records needs rules for which records deserve to exist.

“I used to check every bank app to answer one question: did they pay?”Sebastian Popa / COO and General Counsel

Back in finance, HKR says Jamie checks for payments 14 times daily and gathers 27 account balances into one view. Its paid profit-and-loss dashboard was cancelled only after the replacement matched it line for line. That comparison is a more useful adoption test than a good-looking demo.

05 Buy one role. Measure the change.

The current commercial offer puts an operator inside support, sales, operations, finance or recruiting. During the pilot, the customer pays for the employee; HKR includes the custom AI build. After day 90, it proposes charging per outcome. That creates a different incentive from simply selling another seat.

A buyer can copy the method without copying the whole business: choose one role, establish the baseline, document exceptions and review the numbers before expanding. The fit requires access to existing tools, a team lead and work that can be observed. A buyer seeking a completely hands-off service would still need to supply direction. Highly variable work with little repetition offers less to automate.

HKR sits between staffing and AI implementation: it supplies someone who learns the work and remains responsible after the software arrives. The test is whether the result improves while the responsibilities stay clear. Jamie’s missing permission is a useful place to start.