Summit 2025 Recap #3: The CFO's Transformation Playbook – How to Drive AI Adoption and Cultural Change Across Your Enterprise

Welcome to the CFO Connect Summit 2025 Recap Series - your go-to source for key insights and actionable takeaways from this year's most impactful finance leadership sessions.

On October 8, nearly 500 finance leaders convened to solve a defining challenge: turning "default-to-AI" mandates into operating models that actually improve the P&L. This session, "The CFO's Transformation Playbook: Driving AI Adoption and Cultural Change Across the Enterprise," tackled the hard questions: what proves an executive mandate is real, when to build vs. buy, how to empower teams without spending chaos, how AI changes 2026 hiring plans, and how to think about pricing.

Bottom line: CFOs can turn “default-to-AI” mandates into real P&L impact by embedding AI into the product roadmap, giving teams broad access to safe AI tools, and adopting lightweight governance that encourages experimentation. The most effective finance leaders pair build-vs-buy discipline with agent-driven workflows, redesigned hiring plans, and clear pricing models for scalable AI adoption.

Speaker Credentials

Gabriel Hubert, CEO & Co-founder at Dust – Former Chief Product Officer at Alan and five-year Stripe veteran. Dust enables companies to design, deploy, and maintain AI agents connected to internal data. Firsthand visibility into what drives adoption vs. theatre.

Julien Lafouge, CFO at Photoroom – Former CFO at Spendesk and BlaBlaCar. Photoroom serves global enterprises like Ikea and Amazon. 300 million downloads in 180 countries, with a finance team of three - a team that would be doubled without AI.

Moderator: Jérôme Tomasini, Stripe – Leads Startup & Investor Partnerships at Stripe, where 78% of the Forbes AI 50 rely on Stripe to monetise usage and scale globally.

Key Session Themes & Takeaways

Theme 1: How can CFOs identify a real AI executive mandate (vs. theatre)?

Three signals separate real transformation from talk. 

1. AI in the product roadmap (not just internal productivity).

2. Broad internal access to AI tools (not exec-only or hidden behind IT).

3. Value measured beyond cost savings (decision velocity, ways of working, progress reporting).

This urgency is reflected globally: in early 2024, 72% of organisations reported using some form of AI - up from about 50% the year before - and 65% now use generative AI in at least one business function (The state of AI in early 2024, 2024). Similarly, an IBM survey (2024) of large enterprises found that 42% have already deployed AI, another 40% are actively exploring it, and 59% have accelerated AI investments over the past two years.

Practical Application: Audit your AI agenda against these three signals. If your product roadmap lacks AI or your ROI model is cost-only, you're underinvesting.

Theme 2: What governance model enables AI innovation without creating chaos?

Photoroom's rule: "If it costs <€100 and saves ≥1 hour, don't ask - buy and test." Start in finance. Sum tool costs and estimate productivity gains; if it saves ≥1 FTE and costs less than that FTE, you're winning. Use opportunity cost thinking, and don't force premature ROI - give experiments time to prove. Recent research supports this flexible approach: a late-2024 economic analysis found that 30.1% of workers used generative AI at work, and those who used it for day-to-day tasks reported roughly a threefold increase in task efficiency compared with non-users (Marketing AI Institute, 2025). McKinsey (2023) similarly estimates that generative AI increases the automatable share of knowledge-work activities by more than 30 percentage points in some categories.

Governance Framework: Start with a high-level assessment of your own function first. If AI tools deliver measurable productivity gains in finance (traditionally conservative), you have proof of concept for broader rollout. Use the €100/1-hour heuristic for individual subscriptions while maintaining visibility into aggregate spend and impact.

Theme 3: When should CFOs build AI internally vs. buy existing platforms?

Every company has three AI roadmaps:

• Product AI (customer value): mostly build - it's your moat.

• Automation AI (A–Z workflows): case-by-case.

• Productivity AI (future-of-work tooling): generally buy.

Avoid false economies: a €150K engineer is ~€300K fully loaded. Don't have engineers build internal chatbots you could buy.

Decision Framework: Ask yourself: Will learning from building this compound strengthen our competitive advantage? If not, and it's a day-to-day productivity tool, buy it. If your engineers are building chatbots on retrieval pipelines when you could buy them, you're probably misallocating talent.

Theme 4: Which teams adopt AI agents the fastest, and why does it matter?

Mandate beats function. With clear empowerment, builders emerge across legal, marketing, sales, support, and engineering. Without it, adoption concentrates in engineering (curiosity), support ops, and sales enablement (budget pressure).

Unexpected power users: expert teams (finance, legal, recruiting) - they deploy agents to deflect repetitive questions and reclaim time.

High-adopter persona traits:

• Thinks across teams, not just their own

• Comfortable iterating

• Confident in value beyond specific tasks

Adoption Best Practices: The highest adopters share three characteristics: (1) they think transversely across teams, not just their own function; (2) they're comfortable iterating and experimenting; (3) they're confident about bringing value, not worried about tasks disappearing. These traits correlate strongly with high-performance, high-potential employees you want to retain anyway.

Theme 5: How should CFOs adjust 2026 hiring plans in an AI-driven organisation?

Reality check: transaction-heavy junior roles shrink; strategy roles deepen. Labour-market data reinforces this shift. A 2024 IMF study found that regions with higher AI adoption experienced a larger decline in the employment-to-population ratio, with the impact concentrated in manufacturing, low-skill services, and middle-skill non-STEM roles. Meanwhile, MIT’s 2025 “Iceberg Index” estimates that current AI capabilities are already sufficient to replace roughly 11.7% of U.S. workers, a much higher automation exposure than earlier models predicted. Photoroom's finance team would be 2× larger without AI. Use savings to hire fewer, higher-skill profiles who leverage AI to increase decision velocity and quality. Remember Jevons' Paradox: when something becomes more efficient, you often use more of it to gain an advantage.

The strategic question for CFOs: How much do you value quality increases? How much faster can you make significant strategic decisions? "If it takes three months to make a strategic decision on a new product or enter a new market, and you can do it in two months instead, that's a tremendous market opportunity and strategic advantage versus the competition."

Hiring Framework for 2026: Budget for smaller teams with higher skill profiles, especially in functions with high transaction volume. Invest the savings in premium talent who can leverage AI to increase decision velocity and quality. Don't expect one-to-one headcount replacement – some roles genuinely won't scale the same way, but your total output and quality should increase dramatically.

Theme 6: How should CFOs evaluate AI pricing models across different use cases?

Expect pricing fragmentation:

• Usage-based for power users (e.g., $10K+/mo tokens for elite engineers) - track productivity like marketing spend.

• Per-seat for basic productivity (e.g., €20–40/user/mo) - treat like electricity/G&A.

• Outcome-based for specific, measurable deliverables (benchmarked vs. consultants/agency work).

Practical Guidance for CFOs: Don't expect a single pricing model for all AI tools. Budget different categories separately: (1) power-user consumption for technical roles (track productivity gains, not per-token costs); (2) universal productivity tools as cost-per-employee baseline (like software licenses); (3) project-based or outcome-based pricing for strategic initiatives (compare against consultant/agency costs you'd otherwise pay).

👉 Watch the full discussion

Practical Implications for Finance Leaders

Based on the session insights, CFOs should prioritise these actions:

1. Audit Your Executive Mandate Against the Three Signals: Does your AI strategy appear in your product roadmap? Have you made tools broadly available across teams? Are you measuring transformation beyond cost savings? If you're missing any signal, you're likely not maximising the opportunity.

2. Implement a "€100/1-Hour" Governance Framework: Give teams the freedom to experiment with tools costing no more than €100 that save at least 1 hour. Start by applying this to your own finance function first – if it delivers measurable gains in your conservative environment, you have proof for broader rollout.

3. Clarify Your Three AI Roadmaps: Identify which initiatives are product AI (build), automation AI (depends), and productivity AI (buy). Stop having engineers build internal chatbots when you should be buying horizontal platforms. Calculate the actual fully-loaded cost of build decisions.

4. Budget for Smaller, Higher-Skilled Teams: For transaction-heavy functions, plan for teams 30-50% smaller, with higher per-person compensation. Invest savings in premium talent who can leverage AI to increase decision velocity and quality rather than just processing volume.

5. Fragment Your AI Pricing Budget: Stop applying a single pricing model to all AI tools. Create separate budget categories for power-user consumption (to track productivity ROI), universal productivity tools (per-seat baseline), and outcome-based strategic initiatives (to compare to consultant costs).

6. Measure Agent Adoption by Persona, Not Just Function: Track which individuals adopt agents most heavily. You'll find they're your transverse thinkers, comfortable experimenters, and confident value-creators – exactly the high-performers you most want to retain. Make AI fluency part of performance conversations.

Challenges & Pitfalls to Watch

The speakers identified several common obstacles that derail AI transformation:

Shadow AI Risk: Without enterprise solutions, employees will use unsecured tools with confidential data. Gabriel emphasised: "You do need to roll out tools that make it safe for your employees to try." Don't just mandate adoption – provide secure infrastructure.

Premature ROI Obsession: Both speakers cautioned against rigid ROI tracking in experimental phases. Julien: "We give some time before we actually compute the returns. We don't look at cost first, but at what it can bring for customers and employees." The most valuable applications often have longer-term strategic benefits that aren't immediately quantifiable.

Poor Build Cost Accounting: Gabriel's warning that the €150K employee actually costs €300K fully loaded applies broadly. When evaluating build-versus-buy, include the full cost picture: salaries, benefits, recruiting, onboarding, opportunity costs, and maintenance burden.

Focusing Only on Task Automation: The companies maximising AI value look beyond "doing the thing faster" to "doing the thing better" and "doing things we couldn't do before." Gabriel: "Where CFOs can be very collaborative is thinking about how much faster we can make great decisions. If we can make a strategic decision in two months instead of three, that's a tremendous market opportunity."

Underestimating Maintenance Complexity: The initial enthusiasm around building internal AI tools is normalising as companies realise maintenance is more complex than anticipated. "Some of the ways security and guardrails and actual agent layers need to be thought about are more complex than we thought," Gabriel noted.

Key Quotes & Soundbites

"I'm most excited about leaders trying to maximise upside. They have very ambitious goals, see a lot of growth ahead, and believe this technology is about rewiring how people think and work rather than saving costs in the short term." - Gabriel Hubert, CEO & Co-founder, Dust

"If something costs less than €100 and saves you at least one hour, don't ask – buy it and test it. That's the approach." - Julien Lafouge, CFO, Photoroom

"Five years ago, without the tools we have today, our finance team would be at least double the size. The impact on junior finance roles is real." - Julien Lafouge, CFO, Photoroom

"Because the cost of trying has gone to zero, asking 'why have you not even tried?' is almost a question about your personality, your traits, and your ambition." - Gabriel Hubert, CEO & Co-founder, Dust

"Before asking a colleague, we actually look efficiently using Dust. It's like when I ask my wife by being lazy, 'Where is X, Y, Z?' and she says, 'Just look, and you'll find.' That's the big difference." - Julien Lafouge, CFO, Photoroom

"One engineer on our team spends more than $10,000 in AI tokens per month. They're also among the most senior and productive engineers. We believe it's in the company's interest for them to spend that budget the same way a marketing person spends ad budget." - Gabriel Hubert, CEO & Co-founder, Dust

"We buy the best tools. We build the best tools for our customers. We know we won't do a better job at what Dust is doing than Dust itself." - Julien Lafouge, CFO, Photoroom

Real-World AI Applications Shared

Photoroom's Three-Person Finance Function: Supporting 300 million downloads across 180 countries with enterprise clients including Ikea and Amazon, Photoroom's finance team of three would require double the headcount without AI tools. Julien uses AI for business cases that previously took analyst hours, now completed in seconds.

Dust Engineers' AI Consumption: Individual power users at Dust consume over $10,000/month in AI tokens, yet remain among the most productive team members. This demonstrates the shift from viewing AI as a cost centre to viewing it as a productivity multiplier comparable to marketing spend.

Real-Time Multilingual Support: Gabriel's team replies to Japanese customers in Japanese despite no Japanese speakers on staff, demonstrating quality improvements beyond efficiency gains. This elevates customer experience without corresponding headcount increases.

Automated Expert Deflection: Finance, legal, and recruitment teams deploy agents to answer repetitive questions ("What's our holiday policy? Where's my headcount?"), reclaiming hours previously spent on Slack responding to basic inquiries.

30-Second Business Case Analysis: Photoroom's finance team generates complex business cases in 30 seconds using AI prompts, enabling scenario testing and better decision-making rather than just faster processing of predetermined analyses.

👉 Watch the full discussion

Frequently Asked Questions

When should CFOs build AI tools internally rather than buy platforms?

Building while learning from development compounds your competitive advantage – typically through product AI that differentiates your customer offering. Buy productivity AI for internal workflows, where platforms like Dust offer enterprise-grade security and maintenance at a fraction of the fully-loaded cost of internal development (often €150K salary = €300K total cost). The middle category – automation AI – requires case-by-case assessment based on complexity and strategic value.

How do CFOs measure ROI from AI agent adoption?

Stop creating AI-specific metrics. Instead, measure how AI improves existing KPIs: decision velocity (time from question to strategic decision), expert deflection rates (reduction in basic questions to finance/legal teams), quality improvements (more scenarios tested, better analysis depth), and team output per person. Julien's framework: If AI tools save one FTE's worth of productivity and cost less than that FTE's fully loaded salary, the ROI is clear. For early experiments, give teams time before demanding rigid ROI – the most valuable applications often have longer-term strategic benefits.

What governance model prevents AI spending chaos while enabling innovation?

Implement Photoroom's "€100/1-hour" framework: If a tool costs less than €100 per month and saves at least 1 hour, employees can subscribe without approval. Start by applying this to your own finance function first to validate the approach. Provide enterprise AI licenses to prevent shadow AI security risks. Fragment your budget into three categories: power-user consumption for technical roles (track productivity gains), per-seat productivity baseline (like software licenses), and outcome-based pricing for strategic initiatives (compare to consultant costs).

How should CFOs adjust 2026 hiring plans given AI's impact?

Budget for 30-50% smaller teams in transaction-heavy functions, particularly junior roles focused on data processing and fundamental analysis. Microsoft's stat that 30-35% of code is now AI-generated signals the magnitude. However, invest savings in premium talent with higher per-person compensation who can leverage AI to increase decision velocity and quality. Some roles won't scale the same way, but total output and quality should improve dramatically. Make AI fluency part of hiring profiles and performance reviews.

Which teams typically adopt AI agents fastest, and why does it matter?

With a strong executive mandate, builders emerge across all functions. Without a clear mandate, only engineering (curiosity-driven), support operations, and sales enablement (budget-driven) lead adoption. Surprisingly, expert teams (finance, legal, recruitment) often become power users to deflect repetitive questions. The highest adopters share three traits: transverse thinking across teams, comfort with experimentation, and confidence about bringing value beyond specific tasks. These characteristics correlate with high-performance employees you most want to retain – track adoption by these persona traits, not just function.

How should CFOs think about AI pricing models when costs and value vary so much?

Expect pricing fragmentation, not one model. This shift in pricing mirrors broader enterprise investment patterns. According to McKinsey’s 2024 global AI survey, 67% of organisations expect to increase their AI investment over the next three years, and many industries already allocate more than 5% of their digital budgets to generative AI. Deloitte’s 2024 enterprise research further finds that organisations seeing the most substantial returns focus spending on a few high-impact generative AI use cases and are now testing agentic AI. However, change management and risk governance have become significant constraints. Gabriel identifies three categories: (1) Usage-based for power users – some roles will consume $10K+/month in tokens, comparable to marketing ad spend; (2) Commoditised per-seat for basic productivity tools at €20-40/month, treated like electricity in GNA; (3) Outcome-based for strategic initiatives with measurable business impact, priced against consultant/agency costs you'd otherwise pay. Budget these separately rather than applying a single pricing philosophy to all AI investments.

Recommended Resources & Further Learning

• CFO Connect Pro Community: Join a curated community of finance leaders navigating AI transformation, access exclusive research, and participate in peer learning networks

• Dust Platform: Explore enterprise AI agent deployment with connected internal data and built-in security guardrails

• Stripe for AI Companies: Infrastructure supporting 78% of Forbes AI 50, offering subscription billing, usage-based pricing, and global payment optimisation

• Photoroom for Enterprise: AI-native visual content creation trusted by Fortune 500 companies across 180 countries

Conclusion

The message from this CFO Connect Summit session cuts through the AI hype to reveal a stark competitive reality: execution matters far more than declaration. While headlines trumpet "default-to-AI" mandates, the finance leaders creating genuine competitive advantage are the ones implementing governance frameworks that enable experimentation, making clear build-versus-buy decisions based on strategic value rather than engineering curiosity, and reimagining team structures for a world where AI handles transactions while humans drive strategy.

The opportunity window remains open, but the timeline for advantage is compressing. Companies that moved early – like Photoroom's three-person finance team supporting 300 million downloads, or Dust engineers consuming $10K/month in tokens to achieve outsized productivity – have demonstrated that AI transformation isn't theoretical. It's measurable, it's happening now, and it's creating moats that late adopters will struggle to cross.

For CFOs entering the 2026 planning season, the choice isn't whether to bet on AI – market dynamics have made that decision for them. The choice is whether you'll lead transformation with clear mandates, innovative governance, and strategic resource allocation, or whether you'll manage costs while competitors fundamentally rewire how work gets done.

The 30-day checklist above provides your starting point. The CFO Connect community provides your peer network. The speakers' frameworks provide your roadmap. What remains is execution.

👉 Watch the full discussion