# Bohdan Savchuk

Bohdan Savchuk works in software quality and testing, with a focus on how products behave for real users. His posts also explore security testing for IoT devices and evaluating AI agents on real business tasks.

- Profile: https://b2binfluencer.marketing/creators/bohdan-savchuk/
- LinkedIn: https://www.linkedin.com/in/bsavchuk
- LinkedIn followers: 10,885
- Median of last 8 posts: 16 reactions, 1 comments (checked 2026-11-02)
- Topics: artificial-intelligence, cybersecurity, ai-agents
- Based on 8 dated posts, latest 2026-08-17

## Summary

Bohdan Savchuk is most useful for practical QA, IoT security testing, and evaluating AI agents on real business tasks.

- Best for: A cybersecurity training company brief for a webinar or short course on IoT security and secure QA workflows.
- Not for: An AI-agent platform vendor seeking paid, always-on product advocacy with rates, guaranteed impressions, and disclosure terms.
- Check first: Confirm speaking terms, pricing, deliverables, booking route, and audience composition.

### Takeaways

- Functional tests may miss security risks, and a smart lightbulb can open a corporate network.
- An AI agent that sounds convincing may still make the wrong decision.
- AI agents are something you operate, not just prompt.

## Main ideas

- **A requirement tells you what should happen; a customer shows what actually happens**: Bohdan Savchuk argues that testers who stop once a feature meets its requirements are checking only half the problem. Users arrive with different backgrounds, assumptions, and workflows, so the same product or the same words can mean something entirely different to them. The practical consequence is an extra question before sign-off: not only "Does this work?" but "How might someone else interpret or use this?" Requirements describe intended behaviour; customer perspective is what surfaces actual behaviour.
- **"It worked in the demo" is not enough — and AI agents make that harder**: Bohdan Savchuk's stated focus in software quality is how systems behave when real people use them. With AI agents, that focus widens to include the decisions an agent makes: which information to trust, which action to take, and when to ask for help. An agent can sound convincing and still choose wrongly, so the meaningful test is not whether it performs in a demonstration but how reliably it handles a real business task.
- **Security testing belongs inside quality testing, not beside it**: Bohdan Savchuk argues that traditional QA is not designed to catch IoT vulnerabilities, and that functional testing misses critical risks in connected devices. The proposed correction is a security-first QA workflow: run real attack scenarios through the testing process and integrate security checks without slowing delivery. The underlying position is blunt — in 2026, testing quality without testing security is not really testing.
- **AI agents are something you operate, not just prompt**: Bohdan Savchuk describes the field as past the stage of trying ChatGPT prompts and into chaining agents, orchestrating workflows, and letting them run almost independently. Operating agents that way requires containment as much as ambition: somewhere they can be wrong without consequence. The recurring question in the post is managerial rather than technical — how do you manage a team that never sleeps?
- **Switching off is part of doing the work well, not a reward for it**: Bohdan Savchuk treats rest as a working input. Long-term consistency, focus, and creativity depend on having space to recharge, and staying mentally online around the clock erodes all three. The argument is stated plainly: work will still be there tomorrow, family moments will not.

## Work to read

- [Hacked in Plain Sight: Why IoT Devices Are the New Cyber Battleground](https://www.linkedin.com/posts/bsavchuk_qa-softwaretesting-cybersecurity-activity-7443340162400702465-Gf5V)
- [AI agent evaluation: “It worked in the demo” is not enough](https://www.linkedin.com/posts/bsavchuk_qa-ai-evaluation-activity-7521044195709706240-t9ea)
- [A Mac Mini as a sandbox for AI agents](https://www.linkedin.com/posts/bsavchuk_i-didnt-buy-a-mac-mini-i-bought-a-sandbox-activity-7447007313682493440-H11Z)
- [A T-shirt, a mistaken assumption, and a QA lesson](https://www.linkedin.com/posts/bsavchuk_qualityassurance-softwaretesting-customerexperience-activity-7487188488946073601-X-CU)
- [Judging the 2026 TechForward Awards](https://www.linkedin.com/posts/bsavchuk_techforwardawards-ai-cybersecurity-activity-7471263972839686144-BqQY)

## Brand relationships

- **Anbosoft** (disclosed_partnership)
- **BrowserStack** (disclosed_partnership)
- **SiliconANGLE & theCUBE** (disclosed_partnership)

## Partnerships

Nothing in the material reviewed here sets out a partnership offering. There are no sponsorship packages, no stated rates or deliverables, no disclosure practice, and no inquiry route for commercial work, so a buyer cannot assemble a paid collaboration from what is documented. What the posts do show is Bohdan Savchuk on stage and in the room: a BrowserStack QA Meetup talk in Los Angeles on IoT security testing, a seat on the judging panel for the 2026 SiliconANGLE and theCUBE TechForward Awards, and co-hosting an LA Tech Week basketball evening with Anbosoft that was open to founders, engineering leaders, and AI builders. Those are speaking and community-host appearances rather than a published sponsorship product. The closest thing to a buyer-facing format is a live technical session on secure QA and IoT risk, which matches the brief for a cybersecurity training webinar or course, but the operational terms for booking it are not documented and would need to be confirmed directly.

## Alternatives

- Ian Washburn: Choose Ian Washburn when the training brief shifts from testing connected products to running a vulnerability-management or incident-response curriculum for a university or large enterprise IT audience. The reviewed posts describe supporting the full vulnerability management lifecycle — scanning, validating, prioritizing, and coordinating remediation across endpoint, server, network, identity, cloud, and SaaS environments — alongside SIEM and EDR triage, incident response, and building automated workflows, dashboards, and metrics. A second post frames cybersecurity in higher education as an institutional trust, resilience, and governance function spanning vulnerability management, incident response, security operations, awareness and training, zero trust, AI security and governance, data governance, and policy. That is the raw material for a governance-heavy or security-operations-track course, and it speaks to buyers who need to train staff responsible for protecting institutional data rather than shipping a product.
- Sachin Desai: Choose Sachin Desai when the course needs identity governance, non-human identity, or IAM architecture coverage. The reviewed posts document a named, delivered conference session rather than general interest: “Rethinking Identity Governance: How Unified SaaS and Identity Management Solve Modern Security Challenges,” presented at Identica 2025 in Bengaluru on 6 December 2025. The follow-up summary lays out the argument — identity now includes bots, service accounts, workloads, and AI systems, making their management and governance one of the biggest security challenges — and lists adjacent topics from the event including non-human identities, AI governance, identity risk and hygiene, and the future of IAM architectures. That gives a training company a specific session title, theme, and audience context to build a module around.
- Demawu Apeti: Choose Demawu Apeti when the buyer needs the course anchored in a recognised compliance framework and aimed at industrial or regulated environments. The reviewed AI-governance post argues that adoption must be matched by governance from an ISO 27001 and GRC perspective: integrate AI risk into existing risk management, supplier assurance, change control, and business continuity processes rather than creating governance in isolation, and in industrial settings treat human oversight, accountability, continuous risk assessment, and auditability as especially critical when AI can influence physical operations. A second post records Associate Cyber Security Professional registration with the UK Cyber Security Council and frames it around recognised standards, ethical practice, and continuing professional development. Together these support a compliance-led, UK-credential-linked course.

---
AI-assisted summary of public posts. Audience demographics are not available. Engagement appears only where we checked recent posts. Method: https://b2binfluencer.marketing/methodology/