🎊 AI Spotlight: Anna Kovalova 🎊

We’re excited to present Anna Kovalova, CEO at Anbosoft as this week’s AI Spotlight.

Let’s dive into our interview with Anna and see how she is using AI.

1. Share your AI origin story

I didn’t start in a lab; I started on the test bench. After years shipping software, hardware, and firmware, I began using AI to speed up the “thinking work” in QA—generating test ideas and charters, exploring edge cases, drafting risk notes, and synthesizing logs. Very quickly I learned two truths: AI can accelerate analysis, but quality still depends on human context and judgment.

The pivotal moment came when I saw teams trying AI in isolated pockets with little business impact. What they needed wasn’t another tool—it was a repeatable way to connect quality decisions to outcomes. That insight led me to design an AI-powered QA Audit: a structured survey plus interviews, a maturity score, a risk map, and a prioritized action plan that shows “where you are, where you could be, and how to get there.” I built it to illuminate waste, shorten feedback loops, and translate QA improvements into executive-friendly metrics like cycle time, hotfix frequency, and defect escape.

I shared the approach publicly, then iterated it with real teams across different stacks. Along the way I documented practical, day-to-day AI uses for testers so others could adopt them immediately. That combination—hands-on practice, a measurable framework, and open teaching—defines my AI journey. Today my principle is the same as when I started: keep quality human-centered, apply AI where it truly reduces waste and risk, and package the learnings so any team can benefit.

2. What three AI tools have been most game changing for you?

• ChatGPT — my “Swiss-army” QA workbench

Why it changed my day-to-day: it collapsed the slowest parts of analysis. I use ChatGPT’s data-analysis tooling to ingest logs/CSV dumps, slice defects by signal (owner, env, commit), and spin up quick visualizations—without leaving the chat. That lets me prioritize tests with evidence instead of hunches, then roll straight into generating BDD scenarios, edge-case lists, regexes, or crisp repro steps for non-native readers. The net effect is fewer context switches and faster, evidence-backed decisions.

Why it fits my workflow: the same workspace can reason over screenshots or UI diffs when I need it (multimodal), so I don’t bounce between tools during failure triage.

• Claude — long-context reasoning + tidy “Artifacts”

Why it changed my day-to-day: spec reviews and multi-file plans stopped feeling like a juggling act. Claude’s large context window lets me keep lengthy requirements, prior conversations, and log excerpts “in memory” while I iterate—so I can ask higher-order questions (risk, coverage gaps) without trimming context.

Why it fits my workflow: the Artifacts pane gives me a clean side canvas where test plans, risk tables, or code snippets live as first-class objects I can refine, version, and share—without losing the chat thread that produced them. It turns brainstorming into a tangible deliverable in one place.

• Postman Postbot — API testing on rails

Why it changed my day-to-day: Postbot automates the glue work in API coverage. I describe what I want in plain language and get runnable tests for a single request or an entire collection, plus inline suggestions while I hand-edit scripts. It also helps troubleshoot failing checks and draft documentation where my team already lives—inside Postman. That shortens the path from “new endpoint” to reliable, reviewed coverage.

3. If you were just starting your AI journey today where would you start?

Here’s how I’d start today—based on what actually worked for me, but simplified so a newcomer can get momentum fast (and without breaking anything important).

• Pick one outcome and make it measurable.

I’d choose a single, high-leverage problem (for example: “cut hotfixes,” “shorten triage from hours to minutes,” or “reduce flaky tests”). I’d write the baseline numbers on day one, so every AI change has a clear before/after.

• Set simple guardrails up front.

  • I’d document three rules I’ve learned to be non-negotiable:

  • No sensitive data in prompts.

  • Every AI output has an owner who reviews it.

  • We log prompts/outputs for learning and audits.

  • It takes 20 minutes and saves weeks of cleanup later.

• Assemble a tiny tool kit and templates.

I’d start with the trio I lean on daily: ChatGPT for rapid test ideas, data slicing, and quick refactors; Claude for long-context specs and tidy iteration; Postman Postbot to turn intent into runnable API tests. Then I’d create a few reusable prompt templates: risk-based test ideas, BDD scenarios, log triage, and defect-write-ups.

• Ship one real workflow in a week.

  • I’d take a single API or feature and do an end-to-end pass:

  • Use ChatGPT to outline risks, edge cases, and a small data set.

  • Use Claude to hold the full spec and produce a first draft of the test plan.

  • Use Postbot to generate tests for a collection, add checks, and fix failures.

  • The deliverable isn’t a “demo”; it’s a working slice the team can feel.

• Measure, then iterate.

After that first week, I’d compare results to my baseline: What got faster? What errors disappeared? What still hurts? I’d keep a simple table of prompts that worked, prompts that backfired, and examples that consistently produce good results. This becomes my internal playbook.

• Build a lightweight evaluation habit.

I’d lock in tiny but real checks: pass-rate of generated API assertions, groundedness of summaries against source logs, and a weekly “defect-escape” snapshot. I’ve learned that if you don’t measure, velocity becomes vanity.

• Practice the two muscles that compound over time.

  • Prompt patterns. Role + task + constraints, a couple of crisp examples, and a structured output. I’d practice these on real tickets and logs—not synthetic exercises.

  • Red-teaming your own work. Before I adopt a new AI step, I stress it: missing context, misleading inputs, or ambiguous requirements. That discipline prevents “AI-flavored” bugs.

• Socialize the win and ask for the next pilot.

I’d package the week-one results in a one-pager: the business outcome, the steps, the guardrails, and the measurable deltas. Then I’d ask for a second pilot in a neighboring area (for example, UI checks or log triage), reusing the same scaffolding.

• 30-day starter plan (what I’d literally do):

  • Week 1: Baseline metrics, guardrails doc, tool setup, templates.

  • Week 2: One feature/API from zero to tested; log what worked.

  • Week 3: Add minimal evals; harden prompts; remove manual glue.

  • Week 4: Present results; scale to a second workflow; retire anything that didn’t earn its keep.

That’s the path I wish I had on day one: one outcome, one week, guardrails from the start, and a feedback loop that ties AI to real improvements the team can see.

4. Share the spotlight: Name 3+ women leading in AI we should all follow.

• Mira Murati (Founder & CEO, Thinking Machines Lab; ex-OpenAI CTO) — Now building a new AI lab/product org; valuable for frontier research + team-building lessons.

• “Silicon Valerie” Bertele (Investor & AI/innovation creator) — prolific voice on AI, product, and venture; daily, punchy posts on how teams and founders can use AI practically (plus a window into ecosystem trends).

• Joy Buolamwini (Founder, Algorithmic Justice League) — Champion for algorithmic fairness; follows and shapes the real-world impacts of AI on communities.

5. As a woman in AI, what do you want our allies to know?

Here’s what I want our allies to know—practical, evidence-based, and focused on what actually helps.

Representation is still lopsided—so seats at the table matter. Women remain a minority in core AI pipelines (e.g., only ~18% of authors at leading AI conferences; >80% of AI professors are men). If we’re not in the rooms where datasets, objectives, and guardrails are set, our needs get missed. Invite us in early—on model, data, and policy decisions.

Access and adoption gaps are real. Women are significantly less likely to use generative AI at work, even as usage soars. Close this with equal tooling access, structured time to learn, and visible sponsorship for women experimenting in production workflows.

Bias isn’t abstract—it hurts women of color first. Landmark studies show much higher error rates for darker-skinned women in commercial systems. Allies can push for diverse evaluation sets, bias testing as a release gate, and red-team reviews that include the people most affected.

Inclusion pays off—literally. Diverse leadership teams are more likely to outperform financially. Tie inclusion goals to business metrics and accountability, not just values statements.

Meetings are where inclusion lives or dies. Women are interrupted more and credited less, which quietly shapes who gets stretch work and budget. Use simple norms: no interruptions; “echo + attribute” ideas; rotate facilitation; and capture decisions in writing with owners.

Caregiving bias is a career tax. The motherhood penalty is well-documented. Flexible schedules, outcome-based performance reviews, and normalizing career “zig-zags” keep strong contributors in the field.

AI may reshape women’s jobs more, not less. Roles with high clerical/admin content—where women are over-represented—are especially exposed. Allies in leadership should pair automation with upskilling budgets and clear pathways into higher-value work.

What great allyship looks like in practice:

• Share power (co-own decisions, not just meetings).

• Sponsor, don’t just mentor (open doors to scope, budget, and visibility).

• Codify fairness (structured interviews, skill-based evals, written credit).

• Resource inclusion (time, tools, and training—especially for AI).

• Measure it (publish adoption, promotion, and pay-equity deltas by gender).

  
  

If we do these things consistently—bring women into the design loop, reduce friction to use AI, and enforce everyday inclusion—we get safer systems, better business results, and teams where everyone’s work is visible and valued.

Want to be the next in AI Spotlight? It’s a great opportunity to share your voice with our community! Fill out the WxAI AI Spotlight Nomination Form for your chance to step into the AI Spotlight and to share your voice with the Women X AI community.