In the latest episode of the LCN podcast, we had the privilege of hosting Candy Liu, Strategic Global Account Lead at Microsoft. Candy advises some of Microsoft’s 200 global strategic accounts, which means her role gives her a unique vantage point on what AI adoption really looks like inside the world’s largest enterprises. This blog distils the most important insights from our conversation, but you can watch or listen to the full podcast here. At the end, I have added some concluding thoughts on the practical implications for transfer pricing teams.
1. Enterprise AI Is No Longer Optional – It’s a Strategic Imperative
On this, Candy was unequivocal:
“If a company tells us they don’t have an AI strategy, I think their company is already out in the market.”
Microsoft’s global strategic accounts now treat AI as a board-level priority. Quarterly governance meetings between Microsoft and C-suite leaders increasingly focus on:
- upcoming acquisitions and divestments
- operational bottlenecks
- productivity constraints
- security and governance risks
- where AI can deliver measurable impact
The message is clear: AI is no longer a “technology project”. It is a leading item on the business transformation agenda.
2. The First Step Is Not Custom AI – It’s Getting Every Employee Using AI Daily
Candy emphasised that the biggest misconception about AI is that it begins with complex, bespoke solutions. In reality, the first frontier is behavioural. As she puts it:
“If your staff don’t know how AI can help them, how can you think about AI helping the business?”
Microsoft’s approach is simple:
- Every employee should use AI every day, starting with Copilot embedded in Office.
- Employees should begin creating their own AI agents, small, task-specific assistants that automate repetitive work.
- Only then should organisations prioritise high-impact, high-readiness use cases for custom AI.
This bottom-up adoption is what unlocks the real transformation.
3. AI Agents: The New Workforce Multipliers
One of the most striking insights was Candy’s example from SoftBank:
“Each staff member has 100 AI agents helping them to do their work.”
These agents are not complex systems. They are small, modular assistants that:
- summarise inboxes
- book meetings
- prepare briefing packs
- extract data
- complete expense claims
- generate reports
- automate HR or finance workflows
Crucially, employees can create these agents themselves using natural language, which means no coding is required.
4. Hackathons: The Fastest Way to Discover High-Impact Use Cases
Microsoft runs AI hackathons for organisations of all sizes with a very simple format:
- give employees access to AI tools
- provide dummy or production data
- ask each team to identify a business problem
- build an AI agent or workflow in a few hours
The results are often transformative.
Candy shared one example from Greene King: AI-driven forecasting and automated ordering reduced food waste by £7 million per year, while also delivering major sustainability benefits.
5. What Makes AI Adoption Succeed? Three Patterns Stand Out
Candy has seen both successful and failed AI programmes. The difference comes down to three factors:
a. Executive sponsorship
Without top-down support, AI initiatives stall, especially when they require reallocating resources or upskilling teams.
b. Empowerment, not bottlenecks
AI cannot be routed through IT alone.
“If everything needs to go back to IT, it will fail.”
Teams need autonomy to build and iterate.
c. Identifying “champions” in each function
Hackathons reveal who is naturally curious, passionate and willing to experiment. These champions become the internal engines of adoption.
6. Why Finance and Tax Teams Are Perfect Candidates for AI
Candy highlighted that after customer service, finance is the function where AI has the fastest and most measurable impact.
Why?
Because finance teams spend enormous time on:
- consolidating spreadsheets
- reconciling data
- preparing reports
- forecasting
- validating numbers
- drafting documentation
AI agents can now perform these tasks using natural language instructions.
“You can just tell the agent what you want, and it will generate the outcome, the forecast, the summary.”
For tax and transfer pricing teams, which are often small, under-resourced and dealing with complex data, this is a game-changer.
Candy’s message was clear: AI won’t replace tax professionals. It will elevate them into governance, review and strategic roles.
7. The Future: Agents Working With Agents
Candy offered a glimpse of what’s coming next:
- AI agents will collaborate with each other
- they will take actions, not just answer queries
- employees will manage digital resources the way they manage human teams
- “agent governance” will become a new leadership skill
8. Changing Habits: The Hardest but Most Transformative Step
Perhaps the most practical insight was also the simplest:
“Instead of opening Outlook and your calendar first thing, open your AI chat and ask: what do I need to do today?”
This shift from manually checking information to delegating it is the foundation of AI-enabled productivity.
9. A Final Thought: Just Start
Candy’s closing message was direct:
“Just do it.”
AI is no longer a future technology. It is a present-day competitive advantage. The organisations and individuals who lean in now will define the next decade of productivity.
10. My Thoughts on Why Enterprise AI Matters for Transfer Pricing Teams
For transfer pricing and international tax teams, Candy’s message lands at a critical moment. Across Africa, Europe and the US, TP audits are accelerating, documentation expectations are rising, and DEMPE analyses increasingly require granular, defensible evidence. At the same time, most in-house TP teams are small, stretched, and increasingly managing complex data flows across dozens of jurisdictions.
This is exactly where enterprise AI becomes strategically relevant.
1. AI agents can eliminate the manual burden that undermines TP compliance
TP teams spend disproportionate time on low-value tasks: consolidating spreadsheets, reconciling intercompany data, preparing local files, mapping agreements to actual conduct, and extracting information from ERP systems.
Candy’s point is that AI agents can now perform these tasks using natural language, reliably and at scale, helping in-house teams stay on top of the rising administrative burden.
This frees TP professionals to focus on judgement, governance and defensibility, the parts tax authorities scrutinise most.
2. AI improves the quality and consistency of DEMPE and substance analysis
Our recent webinar with Donal O’Connell highlighted the difficulty of identifying intangibles, trade secrets and operational know-how. AI agents can support this by:
- scanning internal documents for references to processes, know-how and IP
- summarising interview transcripts
- mapping functions to assets and risks
- identifying inconsistencies between agreements, TP files and actual conduct
In a world where “lack of evidence” is the most common reason taxpayers lose TP cases, this is transformative.
3. AI strengthens audit readiness, especially in high-risk jurisdictions
As Cabrini McCarrick noted in our recent webinar on TP landscape in Africa, authorities increasingly demand:
- transaction-by-transaction evidence
- rapid responses (sometimes within 7 days)
- alignment between TP files, financials and legal agreements
- proof of substance
AI agents can pre-prepare audit packs, track documentation gaps, and surface anomalies before an audit begins.
4. AI helps small TP teams operate like large ones
Most in-house TP functions are tiny relative to their responsibilities. AI agents effectively multiply headcount without increasing cost.
For TP teams, this could mean:
- automated benchmarking support
- automated data extraction from ERP systems
- automated variance analysis
- automated drafting of local files
- automated monitoring of intercompany agreements
5. AI adoption is becoming a governance expectation
Candy’s warning is blunt:
“Companies without an AI strategy are already behind.”
Tax authorities are modernising. Multinationals must do the same. AI-enabled TP functions will be better documented, more consistent, more defensible and more resilient.
The Bottom Line
Enterprise AI is not a technology story. It is a governance, compliance and risk-management story, one that directly affects transfer pricing outcomes.
For TP leaders, the question is no longer whether to adopt AI, but how quickly they can embed it into documentation, DEMPE, data management and audit readiness.