EVOLVE connects NED’s, board-level leaders, senior advisors and fractional executives with growth-stage commerce companies. Our events are one expression of that work.

We host expert sessions, in-person meetups, and small, closed dinners across key global markets. These gatherings are designed to support ongoing conversations, not one-off appearances.

Now, onto this week’s newsletter.

At EVOLVE Expert Session 99, we hosted Tomasz Karwatka.

Tomasz co-founded Divante, a European commerce technology firm that later produced ventures including Vue Storefront and Open Loyalty. He subsequently co-founded Callstack and has invested in product-led technology businesses across Europe. His operating experience spans agency scale, product spin-outs, enterprise software delivery and capital formation. “Tom” as he is also called, has built services businesses, products inside services businesses and scaled organizations that lived on billable work.

He understands the internal mechanics of margin, hiring cycles, utilization rates and enterprise procurement.

The opening question was direct:

His response was immediate.

“No.”

The discussion that followed was not rhetorical. It was operational. And it revealed something larger than agency economics.

For two decades, the economics of agencies were straightforward.

Enterprises lacked either speed, skill or capacity. Agencies supplied execution. They billed time. Growth required hiring. Margins were managed through utilization.

This was a linear model.

More revenue required more people.
More people required more cost.

Artificial intelligence disrupts that linearity.

During the session, participants described enterprise deals that would have closed routinely twelve months ago. Instead, they stalled.

One case involved a seven-figure engagement paused because a client built an internal proof of concept using AI tooling. Another involved a banking institution replacing vendor-grade solutions with internally assembled prototypes built with generative coding systems.

These were not speculative experiments. They were procurement decisions.

One participant summarized the shift succinctly:

The significance lies less in the technology than in the confidence. Enterprises are now comfortable experimenting internally with systems that previously required vendor involvement.

McKinsey’s 2023 report The Economic Potential of Generative AI estimates that generative AI could increase software development productivity by 20 to 45 percent depending on task complexity. Goldman Sachs, in its 2023 global economics report, projected that generative AI could expose 300 million full-time roles to automation globally, with knowledge work among the most affected.

The data confirms what operators are experiencing: execution is compressing.

Not universally, but sufficiently to alter procurement behavior.

And procurement behavior determines market structure.

Historically, enterprises outsourced because they lacked internal leverage.

The agency brought:

• Scarce technical skill
• Structured delivery
• External perspective
• Capacity under deadline

AI tooling changes the internal threshold for experimentation.

A product owner without formal engineering training can now assemble a functional prototype. The result may not be elegant. It may not be secure. It may not scale. But it may satisfy immediate internal requirements.

In a budget-constrained environment, that matters.

The Bank for International Settlements noted in its 2024 analysis of AI adoption in financial systems that rapid deployment of AI tools inside institutions shifts operational risk internally if governance frameworks do not mature in parallel. Yet despite such caution, experimentation continues.

Why?

Because the marginal cost of trying has fallen sharply.

Internal experimentation, even if imperfect, often costs less than external procurement. And in uncertain markets, optionality is valuable.

The consequence is subtle but profound.

Agencies are no longer the default starting point for execution.

They are increasingly invited later in the cycle: to harden, secure, integrate or repair.

This alters leverage.

The session returned repeatedly to pricing models.

If code can be generated quickly, billing by the hour becomes fragile.

Tomasz noted that agencies often attempt to respond by repositioning as “strategic.” Yet a 100-person agency cannot simply become a strategy boutique without collapsing its economic base. Large cost structures require throughput.

Tim Gaunt described his own shift toward value-based pricing. The transition was not cosmetic. It required internal recalibration: how projects are scoped, how risk is priced, how sales conversations are framed.

The difficulty lies in reference points.

If a client can produce a proof of concept in a weekend, even if flawed, the starting anchor for pricing shifts downward.

The agency is no longer competing against manual execution. It is competing against a rapidly improving internal toolset.

PwC’s 2024 Global CEO Survey reported that 40 percent of chief executives believe their current business models may not remain viable over the next decade without structural adaptation to AI-driven shifts. Professional services firms, in particular, cited margin compression and pricing pressure as immediate concerns.

The economics are straightforward.

When supply expands dramatically, prices adjust.

Unless value migrates.

Value does not disappear. It relocates.

One of the most important observations during the session was not about technology. It was about proximity.

Agencies sit inside real operational friction daily. They see recurring patterns across clients. They observe problems that internal teams experience as isolated but which, across a portfolio, become structural.

Startups struggle to identify meaningful problems. Agencies encounter them constantly.

This creates informational asymmetry.

The World Economic Forum’s Future of Jobs Report 2023 identifies analytical reasoning, systems thinking and problem framing as among the fastest-growing skill demands. While execution tasks become automatable, defining the right problem remains scarce.

In this environment, leverage shifts toward:

• Architectural coherence
• Cross-project memory
• Risk structuring
• Institutional trust

Those are hard to replicate.

During the session, Tomasz remarked:

In enterprise environments, that is not hyperbole. When liability exposure is high, trust determines who is invited into the room long before output is measured.

A tension emerged that deserves attention.

On one side, enterprises are deploying AI-built systems internally with increasing comfort.

On the other, governance and security concerns persist.

One participant remarked that allowing airport systems to run on unverified AI-generated code would be unthinkable.

Yet banking institutions are experimenting with internal AI builds.

This tension reflects a transitional phase.

Organizations are tolerating fragility in exchange for speed.

The Financial Stability Board has cautioned that AI concentration risk and opaque model dependencies may introduce systemic vulnerabilities in financial systems. Yet internal cost pressures and competitive urgency create incentives to experiment regardless.

The agency is often reintroduced after experimentation reveals complexity.

Instead of leading architecture, agencies may be asked to stabilize systems assembled under speed pressure.

This is not necessarily decline. It is repositioning but it requires recognizing where leverage now sits.

Technology shifts rarely fail because tools are inadequate. They fail because identity resists change.

Tomasz described a case in which a company considered replacing its ERP system. The internal team resisted. The existing system, however flawed, defined their expertise and relevance.

Replacing it threatened status.

AI compresses execution. It also compresses hierarchy.

Managers become supervisors of automated workflows. Engineers become orchestrators rather than builders. Senior operators become bottlenecks if decision velocity does not increase.

Acceleration introduces strain.

Tim noted waking earlier than usual to experiment with AI workflows. He also recognized the risk of burnout as pace accelerates.

Compression increases cognitive load.

The adjustment burden falls disproportionately on mid-career professionals who built identity around scarcity of expertise.

Parallel to agency economics is individual risk.

A widely discussed example involved a senior developer laid off from two high-profile technology companies despite strong credentials. This was not sensational. It was a reminder that employment stability in knowledge sectors is weakening.

Morgan Stanley’s 2024 research on AI exposure in white-collar sectors highlights that productivity gains may not immediately translate into expanded headcount. Firms may absorb efficiency internally.

For individuals, concentration risk increases.

Relying on a single employer becomes structurally riskier than in previous cycles.

This does not imply panic. It implies recalibration.

Board participation, advisory roles, independent consulting and diversified income streams become rational risk management strategies.

The agency conversation and the portfolio career conversation are connected.

Both respond to compression.

Both seek leverage beyond linear employment or linear billing.

The session began with agencies.

It concluded with something broader.

If execution becomes abundant, institutions must reconsider:

• How they price risk
• How they validate trust
• How they structure teams
• How they maintain governance
• How they allocate authority

This is not about extinction. It is about redistribution.

Some agencies will shrink.

Some will productize.

Some will reposition as risk integrators.

Some will disappear.

Enterprises will experiment internally.

Some experiments will fail.

Some will become durable.

The underlying dynamic is not cyclical. It is structural.

Execution is becoming cheaper.

Coordination is becoming more complex..

Three realities emerged clearly.

First, procurement thresholds are falling. Internal experimentation is displacing external vendors earlier in the cycle.

Second, pricing models built on time-based scarcity are fragile under acceleration.

Third, trust networks and proximity to operational friction remain defensible positions.

The conversation was not alarmist. It was practical.

No one suggested agencies vanish overnight.

But no one suggested inertia was viable either.

EVOLVE exists as a working environment for board-level leaders and senior advisors operating across commerce, technology and organizational design.

These sessions are not speculative panels. They are grounded in live operating conditions across markets.

The leverage shifts discussed here are not theoretical. They are visible inside enterprises now.

Membership and corporate partnerships remain selectively open.

Access is ongoing.

The compression of execution will continue.

Where leverage consolidates depends on who understands the system, not who writes the most code.

The question is not whether tools improve.

It is who coordinates when they do.

Curated. Personable. Global.
That is EVOLVE

If you’d like to learn more how to partner with, please email us at
[email protected] or [email protected]

Podcasts & Media We Support