PLUS: Mirakl cuts onboarding 91%, why only 5% of AI projects deliver ROI, and the value stream fix

Morning All, Businesses just unlocked $26B in savings through autonomous spend control systems that detect procurement errors and prevent fraud – with PepsiCo slashing reporting time by 30%.

These measurable wins contrast heavily with enterprise-wide results: 98% of firms deploy AI, yet a mere 5% of projects deliver measurable ROI. Is it time to fundamentally rethink how we implement and track AI's business impact?

Today's dots:

• $26B saved through AI procurement anomalies detection
• Mirakl cuts supplier onboarding by 91% with AI agents
• Why only 5% of AI projects deliver clear ROI
• Value Stream Mapping fixes implementation waste
• Autonomous commerce scales to $2tn spend handling

AI Procurement Tools Save Businesses $26B Through Automated Spend Control

Here's the thing: AI-powered spend management platforms are now autonomously detecting 73% of procurement anomalies while generating $26B in savings, with PepsiCo cutting financial reporting time by 30% through real-time analytics.

Let's unpack that:

• Soldo's AI now flags policy violations and duplicate payments instantly at transaction level, turning spend data into actionable intelligence rather than post-mortem reports
• PepsiCo's partnership with SpendHQ delivered 30% faster financial reporting by mapping all expenditures to real-time dashboards - critical for FMCG supply chains
• Systems automatically categorise 92% of tail-spend transactions (under $5k), freeing procurement teams for strategic supplier negotiations
• Real-time visibility helps detect AI-generated fake receipts, reducing fraud losses by an average 17% across early adopters
• Finance teams using these tools reclaim 80% of monthly reporting time - equivalent to 12 working days annually

If you remember nothing else: These tools transform procurement from historical accounting to forward-looking strategy. The real value isn't just in catching errors, but in shaping purchasing behaviour before money leaves company accounts.

Autonomous Commerce Just Got Real: Mirakl's AI Cut Onboarding by 91%

Here's the thing: Mirakl's AI Catalog Transformer just proved autonomous procurement isn't theoretical – it's slashing supplier onboarding time by 91% while using 25,000 AI agents to manage global inventory. Discover how they did it.

Let's unpack that:

• The system handles 10x more suppliers than legacy tools while reducing categorisation errors by ~50% - meaning fewer lines out of stock and better customer experiences
• It's processing $2tn in spend for top retailers, showing enterprise-ready scalability isn't just theoretical anymore according to Sievo's analysis
• Results mirror wider trends: 34% of procurement teams now say implementing AI/ML is their top tech priority next year per Coupa's research
• The secret sauce? Keeping humans in the loop only for high-judgment decisions while agents handle repetitive coordination tasks 24/7

If you remember nothing else: Mirakl's breakthrough proves AI isn't just optimising procurement - it's fundamentally redefining what's possible. When you train teams to build with AI rather than just use it, you unlock system-level transformation, not just incremental gains.

The AI Maturity Paradox: Everyone's Using It, Very Few Are Winning

Here's the thing: A staggering 98% of professional services firms now deploy AI according to RSM's 2025 survey, yet only 21% effectively measure its financial impact - creating an $18.8B productivity gap for late adopters.

Let's unpack that:

• 91% adoption rates mask shaky implementations - just 1 in 4 companies have AI fully integrated into core workflows (RSM's canvas reveals why)
• Leaders pulling 18% productivity gains use Value Stream Mapping to pinpoint where AI accelerates profit chains, not just task automation
• The ROI crisis starts upstream: 70% of firms need third-party help just to assess AI's impact on operations
• Simple fix: Firms implementing RSM's AI Business Strategy Guide tripled their success rate in linking deployments to P&L improvements
• Hidden risk: Only 36% proactively track regulatory changes despite 85% of data breaches originating from unmonitored AI outputs

If you remember nothing else: Measuring outputs (like documents generated) means nothing without tracking outcomes (revenue per workflow). True AI maturity isn't about adoption rates—it's about operational orchestration.

The Hidden ROI Iceberg in Enterprise AI

Here's the thing: Just 5% of enterprise AI projects deliver measurable ROI according to Aileron's analysis of 765 implementations, while MIT research shows 78% of agentic AI systems fail real-world tasks.

Let's unpack that:

• Most companies aren't tracking financial impact properly - only 21% of AI pilots measured ROI in Aileron's study, making failure almost inevitable without clear success metrics
• Contrary to perception, GenAI experiment costs have dropped 100x since 2023 while capabilities exploded, turning small bets into strategic learning opportunities
• The real risk isn't wasting money on AI - it's losing race-relevant capabilities if recession cuts target the wrong teams while competitors keep building muscle
• Differentiation comes from targeting unique business functions (like supply chain innovation) rather than copying industry-standard use cases in customer service
• Start by making operational leaders - not IT - accountable for AI outcomes, and only build custom solutions after proving value with existing tools

If you remember nothing else: Acting on AI now is like buying Apple stock 50 years ago, even if returns aren't immediate. Waiting risks putting your company at a permanent competitive disadvantage when AI inevitably arrives in your sector - and that catch-up game gets expensive fast.

Cut AI Waste with Value Stream Mapping

Here's the thing: Most AI projects fail because companies automate the wrong things-but Mainsail Partners discovered Value Stream Mapping identifies AI's real ROI zones, helping engineering teams automating QA achieves 50% higher throughput than those focused solely on coding speed. Original framework

Let's unpack that:

• Engineering teams often boost coding speed only to overload QA - VSM reveals AI delivers most value when applied to the real constraint (like automated test generation)
• Customer onboarding saw 91% faster setup at Mirakl through AI that automated data imports and handoffs - no new code required (Mainsail case studies)
• Product leaders should define goals first ("Cut review time by 50%") rather than chasing AI trends - clarity beats hype
• The framework isn't technical: list process steps > time delays > apply AI to worst friction points
• Portfolio companies using VSM report 5x higher AI success rates than industry benchmarks

If you remember nothing else: Only 5% of AI projects deliver ROI because teams automate activities, not constraints. Mapping your value streams first ensures AI accelerates what customers actually pay for—not just what looks shiny.

The Shortlist

RSM reveals 98% of professional services firms deploy AI, yet only 21% effectively measure ROI - creating an $18.8B productivity gap according to their 2025 business survey.

Future reports Grok trails competitors in AI safety metrics, with Anthropic and OpenAI leading in frameworks to prevent malicious use according to their summer 2025 benchmarks.

Brookings proposes consumption taxes as the primary fiscal solution when AI erodes labor income, with autonomous AGI systems possibly requiring harvest-style taxation models in radical scenarios.