The tools arrived first.
The vision arrives now.

A keynote on the AI paradox — and what to do about it.

Every industrial revolution produces the same paradox. The tools arrive first. The vision arrives later.

And in the gap between capability and imagination, entire companies make the wrong bet.

ServiceNow · 2026

→

Average enterprise AI maturity, out of 100.

The map got honest about the territory.

The score fell not from retreat — but from finally grasping the scale of what real adoption demands. Companies stopped counting pilots and started asking what new value those pilots were creating.

The technology arrives

Everywhere but in the productivity statistics.

— Paul David, Stanford · 1990

1879

Edison invents the incandescent lightbulb.

1882

Pearl Street Station — 1st commercial grid.

1890s

Electric motors, commercially viable.

1900

Dynamos everywhere. Productivity flat.

The default, 1900

Swap the power source. Keep the factory.

Vertical

Compact & stacked

Machines clustered around a central shaft. Belts lost energy over distance.

Rigid

Layout frozen

Rearranging any machine meant redesigning the entire mechanical linkage.

Dark

Small windows

Poor ventilation. High injury rates from exposed belts and shafts.

Steam-era factory interior, line shafts and belts

The revolutionary technology was available. Nothing happened.

of U.S. residences used electric lighting in 1900

<5%

of factory mechanical drive was electric in 1900

40yr

before electric motor diffusion hit 50%

The problem wasn't the dynamo. It was what people did with it.

Power sources failed because shafts and belts lost energy over distance, locking machines around the central shaft and freezing the factory's layout.

The factory's architecture was locked to the old technology. Electrification delivered only slight improvements and modest fuel savings, keeping the same layout.

What most asked...

How do we power our existing factory with electricity?

What Ford asked...

What kind of factory would we build if we started from scratch?

Ford's answer

A different factory.

Individual motors

One motor per machine. No shafts, no belts. Each machine independent.

Horizontal layout

Single-story factories organized around material flow — not power.

Flexible workflow

Machines rearranged freely. Lines reconfigured around production logic.

Light & air

Glass walls, skylights. Kahn's "Crystal Palace" design. Safer workers.

Highland Park-era assembly hall, unit drive and material flow

Ford's Factory by the numbers.

Reduced Assembly Time

−87%

728 → 93 min

Cut Model T Price

−70%

$850 → $260

Model Ts Shipped

15M+

Largest manufacturer in the world.

Better factories created better jobs.

$2.34 → $5

Daily wage · Jan 5, 1914

More than doubled.

9hr → 8hr

Shift length

More pay, less time.

32% → 1.4%

Annual turnover

−96%, in two years.

The machine
isn't the constraint.

The real threat isn't falling behind in AI adoption.
It's misunderstanding what AI actually changes.

The world's largest furniture retailer

IKEA 2021, Billie goes live.

An AI chatbot for the predictable tide — delivery status, returns, product availability.

47%

of all inquiries resolved by AI

3.2M+

interactions handled

€13M

in direct cost savings

Most companies would celebrate the cost reduction, lay off the redundant staff, and present the savings to the board. IKEA didn't run that playbook.

What IKEA did instead.

01 · Listened

Studied the 53%

Billie couldn't resolve it — and found customers asking for interior design help. Taste. Spatial reasoning. Creative judgment.

02 · Reskilled

8,500 workers, retrained

Call-center staff became remote interior design consultants. Layered design, sales, and relationship skills on top of existing expertise.

03 · Built new value

45-minute video consults

A human who knew the catalog and could reimagine your space. New revenue line from the same workforce.

What IKEA did instead.

01 · Listened

Studied the 53%

Billie couldn't resolve it — and found customers asking for interior design help. Taste. Spatial reasoning. Creative judgment.

02 · Reskilled

8,500 workers, retrained

Call-center staff became remote interior design consultants. Layered design, sales, and relationship skills on top of existing expertise.

03 · Built new value

45-minute video consults

A human who knew the catalog and could reimagine your space. New revenue line from the same workforce.

What IKEA did instead.

01 · Listened

Studied the 53%

Billie couldn't resolve it — and found customers asking for interior design help. Taste. Spatial reasoning. Creative judgment.

02 · Reskilled

8,500 workers, retrained

Call-center staff became remote interior design consultants. Layered design, sales, and relationship skills on top of existing expertise.

03 · Built new value

45-minute video consults

A human who knew the catalog and could reimagine your space. New revenue line from the same workforce.

Cost saved

€0M

Revenue created · FY22

€0.0B

100× the return of cost-cutting alone.

The easy question...

"What can we eliminate?"

Treats AI as a sharper version of existing tools. Produces real savings. Completely misses the point.

The harder question...

"What can we elevate?"

Sees people, customers, and market position as potential — not fixed inputs. Expands into territory that didn't exist before.

IKEA: €13M saved → €1.3B created.

We have tools that can fundamentally reshape what a company is — and we're using them to make what the company already does slightly cheaper.

That's not transformation.
That's taxidermy — preserving the shape of something that should be evolving.

Taxidermied red fox in a glass display case

Jeremy Waller

So what do you actually do about it?

From diagnosis to practice — architecture, trust, and the right bets.

The pressure to act.

Without a strategy, more activity just deepens the chaos.

73%

of CEOs report stress over AI strategy

64%

fear losing their job over the AI transition

75%

admit their AI strategy is more for show than for guidance

39%

have no formal plan to drive revenue from AI

55%

describe AI use at their company as a chaotic free-for-all

67%

believe their company has suffered a data leak from unapproved AI

Spending accelerated.
Every barrier got worse.

33% → 65%

Difficulty scaling use cases

25% → 62%

Employee skills gaps

34% → 59%

Difficulty quantifying long-term benefit

None of these are solved by better models or more compute.

The Productivity Gap

97% of individuals who use AI report real benefit, but only 29% of organizations see ROI.

48% call AI adoption a massive disappointment.

88%

of executives say their tools are adequate.

21%

of workers agree.

29%

are actively sabotaging the rollout.

Individual wins aren't compounding into business outcomes.
There is a trust gap.

So, what do you do?

From diagnosis to practice — architecture, trust, and the right bets.

Build the right architecture.

Earn trust progressively.

Pick the right bets.

Today's Stack

Software Layer

Applications · APIs · Workflows

Data Layer

Databases · Pipelines · Storage

Infrastructure Layer

Cloud · Compute · Networking · Security

Existing relationships

Intelligence Layer NEW LAYER

Agents · Models · Governed Memory · Orchestration

No vendor lock-in.
You own this layer.

Today's Stack

Software Layer

Applications · APIs · Workflows

Data Layer

Databases · Pipelines · Storage

Infrastructure Layer

Cloud · Compute · Networking · Security

Existing relationships

Intelligence layer relationships (new)

Four pillars of trustworthy AI.

Security

Zero-trust data

The AI only sees what it's authorized to see. Protects people, data, and decisions.

Governance

Reliability

Predictability

Four pillars of trustworthy AI.

Security

Zero-trust data

The AI only sees what it's authorized to see. Protects people, data, and decisions.

Governance

Policy built in

Aligned with business values, compliance, and ethics. Enforcement, not guidelines.

Reliability

Predictability

Four pillars of trustworthy AI.

Security

Zero-trust data

The AI only sees what it's authorized to see. Protects people, data, and decisions.

Governance

Policy built in

Aligned with business values, compliance, and ethics. Enforcement, not guidelines.

Reliability

Auditable results

Cross-validation, confidence scoring, and LLM failover — consistent across runs.

Predictability

Four pillars of trustworthy AI.

Security

Zero-trust data

The AI only sees what it's authorized to see. Protects people, data, and decisions.

Governance

Policy built in

Aligned with business values, compliance, and ethics. Enforcement, not guidelines.

Reliability

Auditable results

Cross-validation, confidence scoring, and LLM failover — consistent across runs.

Predictability

Non-determinism, managed

Human-in-the-loop, observability, and traditional code where it matters.

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

S1 · Probation

Every action reviewed. Human approves all outputs.

Train

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

S1 · Probation

Every action reviewed. Human approves all outputs.

S2 · Supervised

Routine auto-approved. Edge cases flagged for review.

Train → Supervise

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

S1 · Probation

Every action reviewed. Human approves all outputs.

S2 · Supervised

Routine auto-approved. Edge cases flagged for review.

S3 · Trusted

Broader autonomy. Policy-gated on critical actions.

Train → Supervise → Verify

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

S1 · Probation

Every action reviewed. Human approves all outputs.

S2 · Supervised

Routine auto-approved. Edge cases flagged for review.

S3 · Trusted

Broader autonomy. Policy-gated on critical actions.

S4 · Autonomous

Full operational authority. Within defined boundaries.

Train → Supervise → Verify → Expand

Value & Feasibility Framework

Source: Gartner · Customer Service AI Use Case Assessment

Value

Calculated Risks

Likely Wins

Marginal Gains

Quick Wins

Feasibility

The constraint was never the technology.

What is the one process in your business that, if you redesigned it from zero, would change everything downstream?

The tools arrived first. The vision arrives now.

The map got honest about the territory.

Everywhere but in the productivity statistics.

Swap the power source. Keep the factory.

The revolutionary technology was available. Nothing happened.

A different factory.

Ford's Factory by the numbers.

Better factories created better jobs.

IKEA 2021, Billie goes live.

What IKEA did instead.

What IKEA did instead.

What IKEA did instead.

The pressure to act.

Spending accelerated. Every barrier got worse.

So, what do you do?

Four pillars of trustworthy AI.

Four pillars of trustworthy AI.

Four pillars of trustworthy AI.

Four pillars of trustworthy AI.

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

Progressive trust: the apprentice model.

All digital labor starts at zero trust.

Value & Feasibility Framework

The constraint was never the technology.

The tools arrived first.
The vision arrives now.

Spending accelerated.
Every barrier got worse.