This interview is with Olga Kokhan, CEO, Tinkogroup.
For Connectively.us readers, how would you introduce yourself as CEO of Tinkogroup in information services and the specific outcomes your team delivers for clients?
I’m Olga Kokhan, CEO of Tinkogroup — an AI data enablement partner specializing in data annotation, labeling, and validation services that help businesses build better AI products and make smarter decisions.
We work with startups, SMBs, and enterprises across the US and Europe. We handle the data work that most teams don’t have the bandwidth or expertise to do in-house — from image and video annotation to text labeling, data entry, and lead research. Our team delivers 99% accuracy, and we’re fully GDPR-compliant, which matters especially for clients handling sensitive or regulated data.
What sets us apart is how seriously we take quality and communication. Clients don’t just get a vendor — they get a partner who asks the right questions upfront, flags issues early, and keeps them in the loop throughout. That approach has driven real results: last year Tinkogroup grew 3x in revenue, which we attribute directly to repeat business and referrals from clients who trust us with their most critical data projects.
I founded and have been running Tinkogroup for 8+ years, managing everything from business development and client relationships to team structure and delivery quality. For me, data isn’t just a service we sell — it’s the foundation that determines whether an AI model works or fails, whether a business decision is sound or flawed. Getting it right is the only option.
Looking back, which experiences from engineering, data analysis, and hands-on work with tools like AutoCAD and Bluebeam most shaped how you lead Tinkogroup today?
My path to leading Tinkogroup ran through some unexpected places — and I think that’s exactly what shaped my leadership style.
As a Draftsman at JV Poltava Petroleum Company, I worked with AutoCAD in a high-stakes technical environment where precision wasn’t optional. Engineering drawings were either correct or they weren’t. There was no “close enough.” That experience hardwired in me a genuine intolerance for inaccuracy — which is probably why 99% accuracy became a non-negotiable standard at Tinkogroup, not just a marketing claim.
Revision control and documentation work taught me something else: that clean processes protect everyone. When you’re managing complex data projects across distributed teams, the same principle applies. Structure isn’t bureaucracy — it’s how quality survives at scale.
My later work as a Media Analyst at Edelman Intelligence — doing text coding, data analysis, and sentiment analysis — gave me the other half of the picture. I understood data not just as something to be processed, but as something that carries meaning and drives decisions. That perspective is at the core of how I position Tinkogroup: we’re not just moving data, we’re making it usable and trustworthy.
Those two experiences — technical precision and analytical thinking — are the foundation I build on every day as CEO.
As you scaled operations, how did you design your delegation system so judgment stays central while execution scales—share one practice a founder could implement this week?
Scaling without losing judgment is one of the hardest things a founder faces. My approach was structural simplicity: one weekly sync covering every active project, every client, every blocker.
I act as account manager — I stay close to client relationships and strategic decisions, while my Project Manager owns day-to-day execution and is the primary client contact. This separation is deliberate: it means clients always have someone responsive, and I always have visibility without being in every thread.
The practice that made this work was the weekly rhythm. Every Monday, our PMs and I walk through all active projects together — status, risks, anything that needs my judgment or a client-level conversation. Nothing waits longer than a week to surface, and nothing falls through the cracks.
The one thing a founder could implement this week: stop making delegation a case-by-case decision. Define one recurring moment where everything surfaces — and protect it. That single habit replaced dozens of ad hoc check-ins for me and gave my team the clarity to move without waiting for my approval on every step.
Judgment doesn’t scale by being everywhere. It scales by being reliably available at the right moment.
Can you walk us through your verification-first QA workflow at Tinkogroup—closing with one checkpoint a small team could copy tomorrow?
Quality at Tinkogroup is not a final gate — it is built into every layer of the workflow.
-
The first check happens at the individual level: every specialist reviews their own work before it moves anywhere. This sounds basic, but making self-review a formal step — not an assumption — changes how people approach the work from the start.
-
The second layer is cross-checking: a peer reviews the output independently. This catches the errors that self-review misses, because familiarity blinds us to our own mistakes.
-
The third layer scales with the project. For smaller engagements, the PM does a final review before anything reaches the client. For larger or more complex projects, we bring in a dedicated quality specialist. The size of the safety net matches the size of the risk.
On complex projects we also run biweekly team syncs specifically to discuss edge cases — the situations that don’t fit the standard guidelines. These conversations build shared judgment across the team, so everyone handles ambiguous cases consistently, not just the most experienced person in the room.
The one checkpoint a small team could copy tomorrow: make self-review a named, explicit step in your workflow — not implied, not optional. Before anything moves to the next person, the owner signs off on it. That single habit raises the baseline quality of everything that follows.
After adopting your internal workflow dashboard, which single business result improved most, with the measurement approach others could replicate?
The single result that improved the most after introducing structured dashboards was on-time delivery — and the ripple effect was improved client retention.
We run two types of internal tracking at Tinkogroup. Company-wide and department-level scorecards give everyone visibility into performance against key metrics — not just me as CEO, but team leads as well. The second tool is a capacity-planning table that maps client project needs against available team hours in real time. Before we had this, resource allocation was intuitive; after, it became a decision, not a guess.
The measurement was simple: we tracked how often projects were delivered on time and flagged capacity conflicts before they became client problems. When those numbers improved, so did client satisfaction and repeat business.
The approach others could replicate tomorrow: build a single capacity table — clients on one axis, available team hours on the other. Update it weekly. That one view will surface your biggest operational risks faster than any meeting ever could.
Visibility doesn’t just improve execution. It shifts your team from reactive to deliberate — and clients feel that difference even when they can’t name it.
How do you identify and develop Tinkogroup experts—especially among virtual assistants and annotators—so they become quality multipliers for the business?
At Tinkogroup, expert development isn’t a formal program—and I’d argue that’s actually a strength at our stage.
We’re a company in active growth mode, which means we move fast and stay close to our people. Instead of structured career ladders, we pay attention. We watch how someone handles a difficult project, how they respond to feedback, how they treat edge cases. Strong performers reveal themselves—not through performance reviews, but through the quality of their daily work.
When we spot someone with genuine potential, we connect it to a real business need. If the company needs a stronger QA function, and we have an annotator who naturally catches what others miss, that person gets the opportunity. The role grows around their strength, not the other way around.
That’s how quality multipliers emerge at Tinkogroup: not by promoting people up a ladder, but by expanding their impact in the direction they’re already naturally moving.
The one practice others could apply this week is to keep a running mental—or written—note of who handles ambiguity well on your team. That list is your future leadership pipeline, whether you formalize it or not.
When onboarding a new enterprise client with ambiguous specs, how do you run your first-pass calibration to lock scope and quality while keeping communication light?
With AI data projects, locking scope upfront is often the wrong goal. Requirements evolve, edge cases multiply, and what a client thinks they need on day one rarely matches what the model actually requires by week three. We’ve learned to work with that reality rather than against it.
Our calibration always starts with a test task. We take a small, representative sample of the real work and annotate it — not to prove we can do the job, but to generate questions we couldn’t have asked without doing it first. The test surfaces ambiguities that no briefing document ever could.
After the test, we sit down with the client and walk through everything that wasn’t clear. Every grey area gets resolved before it becomes a pattern across thousands of data points. That conversation is where scope actually gets defined — not in the initial call, but in reaction to real output.
This keeps communication light because we’re not asking abstract questions upfront. We’re showing work and asking specific questions. Clients respond to that differently — it’s concrete, it’s fast, and it builds trust immediately.
The one practice to copy: never start with a full rollout. Start with a test, then have the grey-area conversation. That sequence will save you more time than any scoping template ever could.
Which process borrowed from engineering/construction (e.g., revision control or submittal reviews) has most improved data operations quality at Tinkogroup, with a practical way a non-technical team could adopt it?
My first job was as a draftsman at a petroleum company, working with AutoCAD in an environment where a single outdated drawing could cause serious downstream problems. The lesson was simple and permanent: every change gets documented, every version gets tracked, and nobody works from a file they can’t verify is current.
That discipline transferred directly into how we run data operations at Tinkogroup.
When project guidelines change — and in AI annotation work, they often do — we treat it like an engineering revision. The update is documented, versioned, and distributed. The team doesn’t work from memory or from informal Slack messages. They work from the current version of the source of truth, and they know which version that is.
This matters more than it sounds. In annotation projects, an untracked guideline change can corrupt thousands of data points before anyone notices. Revision control makes that problem visible and catchable early.
The practical way a non-technical team could adopt this tomorrow: stop updating shared documents silently. Every change gets a version number, a date, and one line explaining what changed and why. That single habit — applied to any shared workflow document — eliminates an entire category of quality errors that most small teams blame on people rather than process.
Thanks for sharing your knowledge and expertise. Is there anything else you'd like to add?
Building Tinkogroup from a small data services team into a 3x growth AI data enablement partner has taught me one thing above all else: in this industry, trust is the product. Clients don’t just buy annotation or labeling — they hand you the data that their AI systems, their business decisions, and sometimes their competitive advantage depend on. Getting that wrong isn’t an option.
What I’m most proud of isn’t the revenue growth or the Clutch awards — it’s that we’ve maintained 99% accuracy and kept clients coming back while operating through some of the hardest circumstances imaginable, with a remote team based in Ukraine during an ongoing war.
If there’s one thing I’d want Connectively readers to take from my story, it’s this: precision isn’t a personality trait; it’s a system. Build the right checkpoints, hire people who take quality personally, and create a culture where ambiguity gets resolved early — not avoided. That combination scales further than any individual talent ever could.
I’m always open to connecting with founders, operators, and AI teams navigating the data quality challenge. It’s a harder problem than most people realize — and a more solvable one than they fear.