Make the AI opportunity concrete
Translate the deck into a shared view of trends, examples, and implications for Axis.
Axis x Devoteam
From AI pilots to accountable AI value
A meeting brief for discussing where enterprise AI is moving, what the cases show, and why the next wave matters for Axis.
01
Trends
Focused AI, token economics, and governed agent platforms.
02
Cases
GN, Grundfos, Salesforce Agentforce, Maersk, Engie, and Level.
03
Axis discussion
Where AI-enabled APIs and secure data can shape the portfolio.
Meeting intro
What we want to align on
The perspective we want to explore: understand how large end customers - and the surrounding ecosystem - are evolving, and from that decide what offerings and enablers could be valuable to add to the Axis portfolio. Inspiration from comparable Devoteam customer cases helps us see where Axis should focus moving forward.
Move from inspiration to portfolio choices
Discuss which capabilities, services, APIs, data patterns, and operating models could become relevant offers.
Identify the next focused bets
Leave with a short list of topics worth exploring further with business owners and technical stakeholders.
Section 1
Trends we see in enterprise AI
The strongest pattern across the deck is that enterprise AI is becoming less about tool adoption and more about operating discipline: focus, cost control, secure data access, and repeatable delivery.
Trend 01 - HBR signal
Stop running so many AI pilots
HBR's November-December 2025 article makes the point that many companies stay shallow and broad: lots of experiments, weak ownership, and limited P&L impact.
Problem
AI becomes everyone's side project, but nobody owns the transformation.
Shift
Pick fewer areas, go deeper, and build repeatable capability around them.
Metric
Measure business outcomes, not pilot count.
Token cost and AI FinOps move into the boardroom
As AI moves from demo to production, unit economics matter. Every agent, retrieval call, long-context workflow, model choice, and human handover has a cost profile.
- Track cost per task, per interaction, and per outcome.
- Choose the right model tier for the job instead of defaulting to the largest model.
- Design caching, routing, evaluation, and approval flows before scale.
Governed platforms become the fastest route to value
The Salesforce Agentforce part of the deck shows why enterprises prefer building agents where security, access rules, and approved data already exist.
- Agents inherit CRM permissions and field-level access.
- Responses are grounded in approved enterprise knowledge.
- Human handover is designed into high-risk workflows.
Data moves toward AI-accessible architectures
As agents become the main consumers of enterprise data, companies are moving it out of siloed systems of record into governed, query-ready layers that AI can reach safely - the GN-style shift toward a lakehouse such as Snowflake.
- From CRM/ERP silos to a governed, AI-accessible data layer.
- Consolidate into a lakehouse (e.g. Snowflake) agents can query.
- Master data, quality, and security as the precondition for trustworthy agents.
The ecosystem reorganizes around AI-enabled platforms
Value increasingly comes from open platforms that span layers rather than single-vendor lock-in. The Grundfos digital-twin work shows the pattern: intelligent connected components plus simulation, compute, and advisory partners optimizing from component to grid level.
- From proprietary lock-in to open platforms across levels.
- Partners compose: data, models, compute (NVIDIA, Microsoft), and advisory.
- Optimization scales component to plant to grid to ecosystem.
Section 2
Cases that make the trend tangible
The cases show three different kinds of AI value: finding the right opportunities, simulating complex physical systems, and building secure agents on enterprise platforms.
GN Store Nord
AI X-Ray: focus effort toward P&L value
For GN (8,700+ employees, 100+ countries), Devoteam's dual-track approach pairs top-down executive interviews with a bottom-up AI X-Ray that scans the organization for opportunities, then sequences them into a focused, high-value backlog and an agentic operating model.
Top-down executive interviews + bottom-up AI X-Ray
Three lenses: agentic/generative, predictive, economic value
From a long list to a sequenced, P&L-focused backlog
Grundfos
Digital twins from component to ecosystem
The Grundfos examples show AI moving into engineering and infrastructure planning: component-level simulation, plant-level digital twins, and grid-level scenario modeling.
AI and GPU-accelerated CFD in Omniverse
BIM import for plant simulation
Physics-based "what if" planning for water networks
Salesforce Agentforce
Build tailored agents on secured enterprise data
The Agentforce story is about speed and trust. Agents can inherit existing CRM security, use approved data, and hand over to humans when risk or frustration rises.
Security inherited, not rebuilt
Zero-data-retention model routing
Omnichannel service with human handover
Maersk and broader proof points
AI as a production capability, not only a helper tool
Maersk's autonomous coding and Trade & Tariff Studio examples point to a bigger shift: companies can turn specialist knowledge and internal platforms into software-enabled services.
Requirements-layer software creation
Specialist knowledge packaged as an AI platform
Engie and Level show agentic service operations in practice
Section 3
Discussion points: why this is relevant for Axis
Axis can use this discussion to decide where AI becomes part of the customer value proposition, the partner ecosystem, and the technical portfolio.
Where should Axis go deep?
Which customer journeys, partner workflows, service processes, or product experiences have the clearest measurable value?
What data should agents be allowed to use?
How should CRM, product, support, device, and operational data be exposed safely to AI-enabled services?
What should become reusable?
Which APIs, agent patterns, governance controls, integrations, or customer-facing modules could become repeatable portfolio enablers?
How will value and cost be managed?
What AI FinOps model is needed to track token cost, latency, quality, risk, and business impact at scale?
Suggested close:
Pick one or two focused Axis opportunity areas and define what would need to be true for a production-grade AI service.