Module 04V · Lesson 04

Building AI Fluency in Your Organization

Reading time: 18 minutes Track: Role Path — Executive Leadership Prerequisites: Module 04V · Lessons 01-03

What this lesson does

You're not just an AI user. As an executive, you're responsible for whether your organization becomes AI-fluent or doesn't. This is one of the most consequential bets in front of you over the next 24 months.

The MIT data from 2025 — 95% of corporate AI pilots fail to capture value — is not a problem with AI. It's a problem with how organizations deploy AI. The 5% that succeed do specific things differently. Those things are within your control.

By the end of this lesson, you'll be able to:

Identify the organizational moves that distinguish AI-successful biotechs from AI-stalled ones
Recognize the specific failure modes of AI deployment at the organizational level
Make better decisions about AI investment, governance, and culture
Build a deployment plan for your organization

01 · Why this is an executive issue (not an IT issue)

A common pattern: AI gets delegated to IT or to a "head of AI" without executive engagement. The result: technically sound deployments that don't change how work happens.

The reason: organizational adoption of AI requires changes that only executives can make. Specifically:

Resource allocation — investing in the right tools, training, and time for adoption
Cultural permission — making AI use safe, expected, and valued
Governance frameworks — establishing the rules that enable use, not just constrain it
Workflow redesign — restructuring work to take advantage of AI capabilities
Performance expectations — making AI fluency a standard, not optional

Each of these is an executive decision. They can't be delegated. Delegated AI adoption is what produces the 95% failure rate.

02 · The five organizational moves that work

Based on the patterns visible in successful biotech AI deployments:

Move 1 — Executive engagement, modeled visibly

When executives use AI themselves, publicly and visibly, the organization follows. When executives delegate AI to their teams, the organization treats it as optional.

Specifically: senior leaders should be seen using AI in meetings, referring to AI-assisted analyses, and modeling the verification habits they expect of others. The CEO doesn't have to be a power user. They have to be visibly an engaged user.

Move 2 — Specific use case identification

Generic "we're rolling out AI" deployments fail. Specific "here are the 8 workflows we're optimizing in clinical operations, with these specific tools and these expected gains" deployments succeed.

The work: identify, for each function, the 3-8 highest-leverage workflows. Resource the deployment for each. Measure the gains. Iterate.

Move 3 — Workflow redesign, not just tool deployment

Adding AI to existing workflows produces ~20% productivity gains. Redesigning workflows to take advantage of AI produces ~50%+ gains.

The difference: the first deploys AI as a faster version of the current process. The second restructures the process to use AI's strengths.

Example: regulatory section drafting.

Tool deployment: writers use AI to draft sections faster
Workflow redesign: writers spend time on specification and verification; AI does initial drafting; reviewers focus on judgment-heavy revisions; throughput doubles

The redesign requires deliberate work. It doesn't happen automatically.

Move 4 — Governance that enables, not just restricts

Most AI governance defaults to restrictions: don't use these tools, don't input this data, don't do these things. Restrictive governance is necessary but insufficient.

Enabling governance also defines: which tools to use, which data is appropriate where, which use cases are encouraged, what good practice looks like. Enabling governance includes the bright lines (Lesson 1 of Module 3) and the encouragement to use AI broadly within those lines.

The pattern: organizations with only restrictive governance have low AI adoption. Organizations with restrictive + enabling governance have high adoption.

Move 5 — Performance and expectation calibration

When AI is "extra" — something you can use if you have time — adoption stalls. When AI fluency is expected — a normal part of how work happens — adoption sticks.

The work: making AI fluency part of role expectations, performance discussions, and career development. Not as a separate "AI skills" category but as part of being good at the actual work.

This requires manager engagement at every level. Managers who treat AI as optional set their teams to optional adoption.

03 · The five failure modes to avoid

The patterns that produce stalled deployments:

Failure 1 — "Pilot purgatory"

Many companies launch AI pilots that succeed modestly, then never scale beyond the pilot. The pilot demonstrates value; nothing changes about how broader work happens.

Why: Pilots are easier than transformation. Demonstrating value is different from changing how work is done at scale.

Defense: Set pilots up to scale or fail explicitly. Pilots without scaling pathways are demos, not deployments.

Failure 2 — Tool proliferation without governance

Teams adopt different AI tools, often without coordination. Six months later: 12 different AI tools, inconsistent data handling, no organizational learning.

Why: Local adoption is faster than central coordination. The cost of fragmentation is delayed.

Defense: Centralized tool selection with input from functions, deployed broadly enough that local workarounds aren't necessary.

Failure 3 — Training that's content, not capability

Most organizations approach AI training as "let people watch videos about prompting." The content gets consumed; capability doesn't develop.

Why: Capability requires practice. Practice requires structured deployment, not just exposure.

Defense: Training that's tied to specific workflows the trainee actually does. Like this curriculum.

Failure 4 — Risk paralysis

Some organizations respond to AI uncertainty by restricting everything. The over-cautious failure mode (Module 3 Lesson 4) at organizational scale.

Why: Restriction feels safer than enabling. Restriction is also lower-effort than building governance.

Defense: Calibrated governance that enables broadly within clear boundaries. Module 3 specifically.

Failure 5 — Underestimating change management

AI changes how work happens. Changes require change management. Many organizations deploy AI without acknowledging that workflows, roles, and team dynamics will shift.

Why: Tool deployment is concrete; change management is fuzzy. The fuzzy work gets deprioritized.

Defense: Treat AI deployment as change management with tools, not tools with change management.

04 · A deployment framework

A simple framework for thinking about your organization's deployment trajectory.

Phase 1 — Foundation (months 1-3)

Establish governance (bright lines + enabling framework)
Select primary tools
Identify high-leverage use cases per function
Begin executive engagement and modeling
Pilot in 1-2 functions with strong leadership engagement

Phase 2 — Expansion (months 4-9)

Roll out across remaining functions
Build training tied to specific workflows
Track adoption and outcomes
Adjust governance based on what's learned
Build internal expertise (champions, "AI fluency leads")

Phase 3 — Embedding (months 10-18)

Workflow redesign for highest-value functions
Integrate AI fluency into performance frameworks
Develop sophisticated use cases (agents, custom tools)
Build organizational learning loops

Phase 4 — Leverage (months 18+)

Compounding gains from sophisticated use
Competitive differentiation
Talent advantages (AI-fluent organizations attract talent)
Continuous adaptation as AI capabilities evolve

Most biotechs are in Phase 1 or early Phase 2 as of late 2025. The timing of your investment matters; the organizations getting to Phase 3 by mid-2026 will have meaningful advantages.

05 · The "AI fluency lead" role

A specific organizational role that's emerging: someone responsible for AI fluency development across the organization.

This is not the same as a CIO or "head of AI." Those roles focus on tools, data, and infrastructure. The AI fluency lead focuses on capability development across the workforce.

What this role does

Owns the AI fluency curriculum and training
Coordinates between IT (tools) and functions (use cases)
Identifies and develops internal champions
Tracks adoption and outcomes
Owns governance evolution
Reports to senior leadership on AI fluency progress

Where this role sits

In most organizations, this role reports to the COO, CHRO, or CSO. The reporting line matters less than the access to executive attention.

For organizations with <100 people, this can be a part-time responsibility for an existing senior leader. For 100-500 person organizations, this becomes a full-time role. For 500+ person organizations, this is a team.

If your organization doesn't have someone in this role, that's the first hire to consider. The cost is modest; the upside is large.

06 · The investment frame

Thinking about AI investment for your organization:

Tooling costs

For a 100-person biotech, enterprise AI tools cost ~$50-100K/year. Specialized tools may add $20-100K/year. Total: $70-200K/year.

This is small relative to clinical, R&D, or commercial spend. Don't underinvest at this level — it's the leverage layer.

People costs

The AI fluency lead role: ~$200-300K loaded.

Champions / power users in each function: distributed cost (part-time responsibilities for existing staff).

Training time: ~20-40 hours per employee in the first year. At $150K average loaded compensation, this is ~$1,500-3,000 per employee in time investment.

Total people investment for 100-person biotech: ~$400-600K in year one.

Expected returns

A well-deployed AI program produces:

20-40% productivity gains on documentation-heavy work (regulatory, clinical, medical writing)
15-25% on operational work (clinical ops, vendor management, project management)
10-15% on analytical work (where the human judgment is the bottleneck)

For a 100-person biotech, this is the equivalent of 15-30 FTE-years of capacity per year. At $150K per FTE, that's $2-5M of effective additional capacity.

Return on investment: 5-10x in capacity gains alone. Plus competitive advantage, decision quality, and talent attraction effects.

This math is real for organizations that deploy well. It's not real for organizations stuck in pilot purgatory.

07 · The "build vs. buy" framing

A common question: should we use commercial AI tools or build our own?

For most biotechs, the answer is: use commercial. Building requires:

ML engineering talent (scarce, expensive)
Infrastructure
Ongoing maintenance and capability investment
Time to capability

Commercial tools provide frontier capability immediately. The capability gap between commercial and homegrown will only widen.

Exceptions where building makes sense:

Highly proprietary data that can't enter commercial environments
Specific workflows where commercial offerings fundamentally don't fit
Strategic differentiation where the AI capability itself is the product

Where building usually doesn't make sense:

General productivity gains
Most documentation workflows
Most analytical workflows
Most communication work

The right architecture: commercial frontier AI as the foundation, with specific custom integrations (MCP servers, custom workflows) for organization-specific needs.

08 · What you can do this week

Specific actions for an executive reading this lesson:

This week

Audit your own AI usage. Are you engaged or delegating?
Identify your organization's current phase (1-4 above)
Have one direct conversation with your senior team about AI deployment status

This month

Identify or designate your AI fluency lead (or interim equivalent)
Verify governance is calibrated (enabling, not just restrictive)
Identify the top 3 workflows for organizational redesign

This quarter

Implement structured training for at least one function
Establish measurement of AI adoption and outcomes
Make AI fluency expectations explicit in performance discussions

This year

Move from Phase 1-2 to Phase 3
Build competitive differentiation in at least one function
Position your organization for the compounding gains of sophisticated AI use

09 · Knowledge check

Three questions.

Q1. Why is AI adoption described as fundamentally an executive issue, not an IT issue?

a) IT can't handle AI b) Organizational adoption requires changes only executives can make — resource allocation, cultural permission, governance frameworks, workflow redesign, performance expectations — delegating these produces the 95% failure rate c) AI is too technical for IT d) Executives should make all decisions

Q2. What's the difference between tool deployment and workflow redesign?

a) Tool deployment is cheaper b) Tool deployment adds AI to existing workflows (~20% gains); workflow redesign restructures processes to use AI's strengths (~50%+ gains) — the redesign requires deliberate work that doesn't happen automatically c) Workflow redesign is too slow d) Tool deployment requires more training

Q3. What's the typical ROI math for a well-deployed AI program at a 100-person biotech?

a) 1-2x investment in productivity gains b) Approximately 5-10x in capacity gains alone — $400-600K total investment in year one producing $2-5M of effective additional capacity, before competitive advantage, decision quality, and talent effects c) Negative ROI for two years d) Cannot be calculated

Answers: Q1: b · Q2: b · Q3: b

10 · What's next

Lesson 05: Capstone — your executive AI playbook.

End of Lesson 04.