Director of Technology Program Management says AI without sufficient context can result in dangerous decisions
Discover how a technology program leader uses AI to reduce reporting overhead, surface delivery risks, and help teams focus on decisions instead of administration.
Anika Diachuk
Director of Technology Program Management at ACV Auctions
Human-Centric Leadership: Anika emphasizes that effective AI adoption starts with empowering teams, not just with technology.
Efficiency Gains: AI tools have streamlined repetitive tasks, improving reporting efficiency and delivery visibility across teams.
Prevent Over-Reliance: Human intervention is essential to interpret AI outputs accurately and ensure business context is maintained.
Simplify Workflows: Over-engineering can hinder clarity; focusing on high-value decisions facilitates faster adoption of AI tools.
Ownership Matters: Clear decision ownership enhances AI integration, helping organizations avoid confusion and improve outcomes.
Anika Diachuk is the Director of Technology Program Management at ACV Auctions. She has led multiple globally distributed, cross-functional delivery teams. And she now focuses on making AI adoption non-intrusive, while making delivery more effective.
We caught up with Anika to find out how she's implementing AI in her delivery workflows. Here's what she said.
Human-centric AI leadership
I am a Ukrainian-native Canadian who gets stuff done. Most people assume my path into AI-driven transformation started with technology, but it started with people — recruiting talent, building trust across global teams, and learning how organizations behave under pressure.
Unlock for Free
Create a free account to finish this piece and join a community of forward-thinking leaders unlocking tools, playbooks, and insights for thriving in the age of AI.
Have an account? Log In
I have now led multiple globally distributed, cross-functional delivery teams supporting enterprise-scale digital marketplace and operational transformation initiatives. Sizes varied from six people to a few thousand. Most of these organizations moved fast — running agile delivery across product, engineering, ops, and business stakeholders while managing complex integrations, rapid scaling, and data-driven platform work for clients spanning automotive dealers, banks, retailers, nonprofits, and global enterprise.
Leading delivery at different organizations taught me that the biggest barrier to innovation is rarely the technology itself; it’s whether people feel empowered enough to change how they work. That’s why my perspective on AI leadership is unconventional: The future of delivery won’t belong to leaders who automate the most, but to those who can make AI feel deeply human, collaborative, and creatively energizing for the teams using it every day.
How AI tools increase workplace efficiency
In the past year, I've replaced the most repetitive, predictable tasks in the delivery process, freeing up time for the true work — mostly by embedding AI-driven summarization and prioritization into status reporting, dependency tracking, and stakeholder communication processes.
As a result of AI, I've seen improved reporting efficiency, delivery visibility, and risk identification across distributed teams, while reducing manual coordination effort and improving execution predictability.
As a result of AI, I've seen improved reporting efficiency, delivery visibility, and risk identification across distributed teams.
But while I can rely on AI to accelerate data-heavy delivery activities such as prioritization insights, status summarization, risk pattern identification, dependency tracking, and operational reporting, I can't rely on it for everything.
Decisions involving stakeholder alignment, team dynamics, organizational change, conflict resolution, and strategic trade-offs remain intentionally human-led. Because successful delivery still depends on trust, context, empathy, and judgment that AI cannot fully replicate.
Why AI outputs require human intervention
In fact, my biggest challenge has been preventing teams from over-relying on automated outputs without applying business context. To prevent teams from doing this, I challenged their outputs and stressed the importance of hands-on experience with product and business context. It's the difference between good and great.
Human intervention is crucial in prompting and interpretation. You need a solid, structured request with correct correlation and causation context, plus a deep understanding of processes and drivers to interpret the outputs.
For example, a tool can tell you that the deliverable is at risk, but the cause might be that the engineer has a sick kid, so you need to dig deep to understand the rationale behind the data and the human behind the task.
My biggest challenge has been preventing teams from over-relying on automated outputs without applying business context. Human intervention is crucial in prompting and interpretation.
How over-engineering hurts workflows
As far as tooling, we currently use Omni for dashboards, Rovo on Jira for engineering management, and internal AI chat for team updates and summarization.
Early on, we could have avoided over-engineering the workflow with too many signals and dashboards. It created more interpretation overhead than clarity. A simpler, tightly scoped use case — focused on one or two high-value decisions — would have driven faster adoption and less skepticism from teams.
Why AI fails at improving decision speed
AI didn’t deliver the expected impact of improving cross-team alignment and decision speed simply by adding more automated insights and dashboards.
Yes, AI improved visibility, but it didn’t reduce ambiguity or resolve conflicting priorities — in some cases, it amplified noise, with different teams interpreting the same outputs differently.
AI improved visibility, but it didn’t reduce ambiguity or resolve conflicting priorities — in some cases, it amplified noise.
Take a dashboard as an example. Each team focuses on its own KPIs and business needs, so they can prompt the tooling to generate output that serves their own key objectives. Building cohesion in this chaos is critical, considering the impact on all teams.
Why better planning doesn't necessarily lead to better outcomes
AI made me let go of the assumption that better planning automatically leads to better delivery outcomes. Even highly detailed plans become less predictive once AI surfaces the real pace of change, and shifts priorities and dependency volatility in real time.
Continuously re-planning based on live signals, rather than more upfront certainty, improved outcomes. Delivery stopped being about getting the plan right and became about adjusting it quickly.
Our AI-powered delivery workflow ingests work items and team updates, where AI clusters themes, flags risks, and auto-summarizes progress across squads. Those insights feed a human-led prioritization session, where decisions become sprint plans. During execution, AI continuously highlights deviations and emerging issues.
The net effect was less time reporting, more time fixing delivery problems.
Why status reporting needs to be redesigned with AI
One area more delivery leaders should actively redesign is status reporting and delivery governance — it’s still one of the most manual, low-value, and distortion-prone parts of most organizations.
I am working on replacing traditional status decks and manual rollups with an AI-augmented workflow that continuously synthesizes updates from delivery tools, flags risks, and generates concise “decision-ready” summaries for leadership. The shift wasn’t just efficiency — it changed the nature of governance conversations from “what’s the status?” to “what decisions do we need to make now?”
The result was less time spent compiling updates and more time resolving actual delivery constraints, with faster escalation of real risks—but only after we tightened decision ownership so AI outputs didn’t become another layer of interpretation.
Why clear ownership is the real unlock with AI
Here's my advice:
Don’t confuse more AI output with better decisions; it just gives people more things to disagree about. The real unlock is clear ownership. AI suggests, humans decide, no committee drama.
Also, keep this in mind. If adoption feels slow, it’s usually not the model — it’s the humans politely ignoring it.
Follow along
You can follow Anika Diachuk's work on LinkedIn.
More expert interviews to come on The Digital Project Manager.
