From retrospective analysis to enterprise intelligence: How HPE CFO Marie Myers is rewiring finance with AI
AI has been "promising" change since the Deep Blue era. Marie Myers, CFO of Hewlett Packard Enterprise, says finance is still early - and that's exactly why she's accelerating.
Her path runs through Compaq and HP's controller ranks to a first CFO tour at UiPath. That's where she learned automation alone won't fix finance. The next leap is agentic AI plus GenAI, deployed with governance, and owned by finance.
HPE's position: infrastructure at the center of AI
HPE sits in the heart of the data center - servers, storage, networking. That's where AI runs at scale. Yes, they compete with Dell, but the point is clear: the infrastructure you build shapes the intelligence you can use.
This is why Myers is pushing hard. The stack matters, and finance needs a seat at the architecture table.
Why a CFO is leading the charge
The CFO allocates capital and sets operating tempo. That makes the CFO the natural steward of AI priorities, guardrails and value capture.
"You can't be a CFO today and not be literate on AI."
From 100-slide decks to live decisions
HPE's weekly 90-minute operational review used to demand ~100 PowerPoint slides pulled together over a weekend. Useful history, slow decisions.
GenAI flipped that. The team now surfaces live insights, reduces prep overhead, and shifts the conversation from "what happened" to "what will we do next." Finance moves from reporter to operator.
The stack: agentic + GenAI, grounded in enterprise infrastructure
Myers is co-developing CFO Insights with Deloitte - a solution that blends agentic AI and generative AI on HPE's Private Cloud AI. The stated aim: a transformation with a purpose - from retrospective performance analysis to forward-looking enterprise intelligence.
The takeaway for finance leaders: pair agents (to run workflows and decisions) with GenAI (to synthesize and communicate) on infrastructure you can control. See HPE's AI approach for context: HPE AI.
Accuracy and control: one number, every time
Finance can't tolerate "it depends" answers. If revenue is 99.9, it must be 99.9 - no matter who asks the model.
To tighten accuracy, HPE worked with NVIDIA to build and run models in a governed environment. More context here: NVIDIA AI Enterprise.
- Standardize data definitions and lineage before model work.
- Constrain retrieval to approved sources; log prompts and outputs.
- Version models, prompts and datasets; enforce change controls.
- Use evaluation suites that test precision, consistency and drift.
- Keep humans in the loop for material decisions and disclosures.
Workflows first, then AI
Agentic AI exposes messy processes. You can't automate a sloppy workflow and expect clarity.
HPE is defining what agents do and what humans own, then rebuilding processes so agents can run reliably. That sequencing matters more than the model choice.
Reskilling everyone (not just a tiger team)
Myers is upskilling more than 3,000 people across and adjacent to finance. The mandate: every role learns how to design workflows, build/use agents and measure impact.
The lesson from RPA: projects fail when change management is underfunded. Budget time, training and incentives - or budget for rework.
A practical playbook for CFOs
- Set a literacy baseline: principles of GenAI, agents, RAG, privacy and evaluation.
- Pick high-frequency, decision-heavy use cases: operational reviews, forecasting, close, variance analysis, cash risk.
- Design for "what we'll do," not "what happened": prompts and dashboards must end in actions and owners.
- Stand up a governed data layer: definitions, access, lineage and policy before model access.
- Adopt agent patterns: intake → retrieve → reason → act → log → escalate.
- Instrument ROI: prep hours removed, decision latency, forecast error, cycle time, working capital.
- Create dual tracks: quick wins in operations, deeper builds for planning and scenario engines.
- Publish rules of the road: model usage, sensitive data, disclosure, and audit trails.
Near-term wins to target
- Operational reviews: auto-briefs, anomaly calls, recommended actions with owners and deadlines.
- Forecasting: scenario narratives, range reasoning and explainability with live driver inputs.
- Close and controllership: variance narratives, policy checks, attachment extraction, control logs.
- Working capital: agent-led collections outreach, dispute triage, and cash risk signals.
- Spend and procurement: supplier summaries, contract deltas, compliance flags, and savings tracking.
The honest status check
Myers is clear: we're in the early innings. The real impact on finance is still ahead.
That's the opportunity. Teams that build literacy, fix workflows, and move fast with guardrails will bank the gains first.
Next step
If you want a structured way to upskill the leadership team and your analysts, start here: AI Learning Path for CFOs.
Your membership also unlocks:
