Archie D'Souza
Walk into almost any project management conference or
software demo today, and you will hear two competing extremes. On one side, technical
deep dives quickly descend into model architectures, vector spaces, and
training weights that feel far removed from daily delivery. On the other side,
glossy product pitches promise "effortless AI magic" that skips over
how the system actually behaves when real-world projects get messy.
I’ve been a project leader for close to four decades and have
realised that you do not need to earn a machine learning degree or write Python
scripts to lead effectively in an AI-enabled environment. Just as a senior
operations or logistics manager does not need to rebuild an aircraft’s avionics
or decode raw flight-data telemetry by hand to make sound operational
decisions, you do not need to become a data scientist to govern AI on your
projects.
What you do need is minimum viable technical literacy:
a practical, grounded grasp of how inputs shape outputs, where models can
mislead, and how to ask sharp, accountable questions.
1.
Ask Vendors What Happens Beyond the Happy
Path
Software demos are built to showcase ideal conditions—clean
data, neat timelines, and frictionless workflows. In real project delivery,
data is incomplete, priorities shift mid-sprint, and exceptions happen every
day.
When evaluating or rolling out an AI-powered project tool,
look past surface features and ask practical questions such as:
- Data
lineage: Which project systems, documents, or communication channels
does the tool actually ingest?
- Model
currency and scope: Does it rely on generic patterns, my
organization's historical projects, or real-time updates?
- Failure
modes: What happens when inputs are sparse, contradictory, or
outdated—does the system flag low confidence, or present a guess with
false certainty?
If a tool cannot explain its assumptions and limitations in
plain language, adopting it introduces silent governance risk.
2.
Guard Data Hygiene Like a Core
Deliverable
AI models do not experience reality directly; they see the
digital breadcrumbs left behind in task boards, schedules, status notes, and
logs. When project data is neglected, even an intelligent system simply scales
up bad assumptions faster.
Keep an eye out for everyday data hygiene red flags:
- Tasks
left in outdated statuses long after work shifts or stalls.
- Vague,
inconsistent, or duplicate work items that blur dependencies.
- Critical
decisions or scope changes living in side chats instead of shared records.
- Teams
gaming metrics or status colors just to keep dashboards looking green.
Strong data discipline isn't glamorous busywork—it is the
foundation that keeps AI insights anchored in reality.
3.
Insist on Explainable AI (XAI) in
Governance Reviews
In a steering committee, milestone review, or risk
escalation, "the AI flagged it" is never a complete rationale.
Project leadership requires accountability, trade-off analysis, and sound
judgment.
That is why Explainable AI (XAI) matters so much in
practice. When a platform forecasts a slip, highlights a cost risk, or
recommends shifting resources, treat the output as an analytical hypothesis to
test—not an unchallengeable verdict. Ask what signals drove the insight, what
assumptions sit underneath it, and whether human context changes the picture.
The Sweet Spot of Leadership
You don't need to master the math behind the screen to stay
in command of the project. By building minimum viable technical literacy,
protecting data quality, and demanding clear explanations during key decisions,
project leaders can avoid both blind trust and knee-jerk scepticism—turning AI
into a genuinely dependable partner in delivery.
Would you like to tailor this with a stronger logistics or
project governance angle, or use it as a template for another post in your
series?