Most organizations still plan in fragments. Finance builds the budget in one tool, sales plans quota and pipeline in the CRM, HR models headcount in a workforce system, and supply chain runs demand and supply planning somewhere else. Each plan is internally consistent and collectively contradictory. When the sales forecast moves, it can take weeks for the impact to reach the cash plan, by which time conditions have changed again.
Extended planning and analysis, usually shortened to xP&A, is the response: extend the discipline and tooling of financial planning across the operational functions so that one change ripples through every connected plan. The idea is not new, but in 2025 and 2026 three developments made it far more practical: AI agents embedded in planning platforms, zero-copy data sharing between ERP systems and data platforms, and a growing expectation from CFOs that finance should orchestrate decisions across the business rather than consolidate spreadsheets at month end. This article explains what xP&A means today, what has changed, and how to approach it without a multi-year big-bang program.
What xP&A Actually Means
At its core, xP&A is a design principle: plans across functions share common data, common drivers and a common calendar, and they are linked so that assumptions flow between them. A practical definition has four components.
- Shared drivers. Volumes, prices, headcount, productivity and capacity are defined once and used by every plan that depends on them.
- Linked models. The sales plan feeds the revenue plan; the workforce plan feeds operating expenses; the supply plan feeds working capital and cash. Changes propagate automatically or on a defined cycle.
- One version of actuals. All plans compare against the same reconciled actuals, typically from the ERP and a governed data platform.
- A common planning cadence. Functions plan on aligned cycles, increasingly rolling rather than annual, so that scenarios can be evaluated across the business at the same time.
xP&A is not one giant model owned by finance. In practice, each function retains ownership of its plan and its operational detail, while finance owns the integration points, the financial translation and the governance of shared drivers. Workday’s 2026 FP&A outlook describes this as FP&A becoming a central orchestrator of data: when HR changes a headcount plan or sales updates a pipeline stage, the cash impact shows up in the planning model.
Why xP&A Is Gaining Momentum Now
AI agents arrived inside planning tools
The most visible change is that planning vendors now ship agents rather than just chat assistants. Pigment, for example, announced in September 2025 an Analyst Agent that continuously monitors data, flags anomalies and risks, drills into root causes and drafts reports and summaries, with Modeler and Planner agents announced to build and adapt models and simulate scenarios. Workday positions agentic AI as going mainstream in planning, including agents that draft first-pass variance commentary and generate driver-based scenarios. SAP has been turning Joule into a platform of role-based agents across its Business Suite and said in October 2025 that it had added more than 40 agents.
Agents matter for xP&A because the hard part of connected planning has always been the work between plans: reconciling definitions, chasing updates, explaining variances and rerunning scenarios. That is exactly the kind of multi-step, cross-system work agents are designed to take on.
Zero-copy data sharing reduced integration friction
Connected planning depends on timely, consistent data from ERP, CRM, HR and supply chain systems. Traditionally that meant fragile extract pipelines. Vendors are now emphasizing direct sharing. SAP’s Business Data Cloud Connect, for instance, shares SAP data with platforms such as Databricks without traditional ETL, with Google BigQuery support planned for the first half of 2026, according to SiliconANGLE’s coverage. Similar patterns exist across Snowflake, Databricks and Microsoft Fabric. For planning teams, this means operational detail can be available in the data platform faster and with less custom engineering.
CFO expectations shifted from adoption to impact
Gartner’s 2025 AI in finance survey of 183 finance leaders found that 59 percent said their finance function used AI, almost unchanged from 58 percent in 2024, and that 91 percent reported low or moderate initial impact. The top implemented use cases were knowledge management, accounts payable automation and anomaly detection, not planning. The message for planning leaders is that experimenting with AI is no longer differentiating; connecting it to decisions that move the business is. xP&A provides that connection.
xP&A is less about buying a bigger planning tool and more about agreeing on shared drivers, linking plans at a few critical points and letting AI handle the reconciliation work that used to consume the planning cycle.
The xP&A Capability Map
When assessing readiness, it helps to break xP&A into capabilities rather than products.
- Data foundation. Reconciled actuals and operational detail in a governed platform, with consistent master data for products, customers, cost centers and locations.
- Driver library. A documented set of business drivers with owners, definitions and the plans that consume them.
- Linked planning models. Functional models connected at defined integration points, with clear direction of flow and timing.
- Scenario engine. The ability to change drivers and see the effect on P&L, balance sheet, cash and operational KPIs together.
- Forecasting and analytics. Statistical or machine learning baselines for key drivers, plus variance analysis against actuals.
- Agentic assistance. Agents that monitor, explain, draft and simulate, within governed permissions.
- Governance and workflow. Version control, approvals, audit trails and clear ownership for each plan and driver.
Why the driver library comes first
Of these capabilities, the driver library is the one most often skipped and most often responsible for failure. A driver library is simply a governed list of the variables that move the business, each with a precise definition, a unit, a grain (for example, by product family and month), an owner, a source of actuals and the list of plans that consume it. Building it forces conversations that fragmented planning avoids: whether sales and finance mean the same thing by “bookings,” whether headcount means full-time equivalents or people, and who is allowed to change the price assumption. It also becomes the context that AI agents need. An agent asked to explain a margin variance can only do so reliably if it knows which drivers feed margin and how they are defined.
Where AI Agents Fit, and Where They Should Not
Agentic AI in planning is still early, and the most productive deployments assign agents clear roles with human accountability.
Good fits today
- Monitoring and alerting. Watching actuals and leading indicators daily, flagging deviations from plan before the month closes.
- Variance explanation drafts. Decomposing variances into price, volume, mix and timing, and drafting commentary for analysts to edit.
- Data preparation. Mapping new cost centers, reconciling hierarchies and detecting missing or anomalous inputs from functional plans.
- Scenario generation. Producing a set of driver-based scenarios around a question such as a supplier price increase, for planners to refine.
- Narrative and reporting. Turning approved numbers into board-ready summaries and slides.
Keep humans firmly in charge of
- Setting targets and approving plans.
- Changing driver definitions or model logic in production.
- Interpreting strategic trade-offs between functions.
- Any output that will be published externally or used for guidance.
A Pragmatic Roadmap to xP&A
Big-bang xP&A programs that try to connect every function in one release tend to stall. A staged roadmap works better.
Phase 1: Prove one critical link (8–12 weeks)
Pick the integration point that causes the most pain. For many companies it is sales to revenue: the pipeline and quota plan in the CRM does not match the revenue forecast in finance. Agree on shared drivers (bookings, conversion rates, average deal size, ramp), connect the two plans, and run one forecast cycle on the linked model. Measure cycle time and forecast accuracy before and after.
Phase 2: Add workforce and operating expense
Headcount is usually the largest controllable cost, and workforce plans change frequently. Link the HR plan to operating expenses so that hiring changes update cost forecasts automatically, with salary, benefit and timing assumptions owned jointly by HR and finance.
Phase 3: Connect supply chain and working capital
For product companies, link demand and supply plans to inventory, purchasing and cash. This is the most complex link because operational plans are granular and change often, so define the aggregation level and refresh cadence carefully.
Phase 4: Introduce agents and rolling cadence
Once links are stable and data is trusted, introduce agents for monitoring, variance explanation and scenario generation, and move from an annual budget to a rolling forecast cadence. Agents work best when they operate on connected, well-defined data; deploying them earlier tends to automate confusion.
Worked Example: A Price Increase That Touches Every Plan
Consider a consumer goods manufacturer with plants in two regions. A key packaging supplier announces a significant price increase effective next quarter.
In a fragmented setup, procurement updates its cost model, finance hears about it in the next monthly review, sales continues to plan promotions on old margins, and supply chain keeps the existing order schedule. The impact on gross margin and cash becomes visible only when actuals arrive.
In an xP&A setup, procurement updates the packaging cost driver in the shared driver library. A monitoring agent detects the change, identifies every product and plan that uses the driver, and drafts three scenarios: absorb the cost, pass through part of it with a list price change, or switch volume to an alternative supplier with a longer lead time. For each scenario, the linked models show the effect on product margins, the promotional plan, inventory levels and the 18-month cash forecast. Sales planners review the price elasticity assumptions, supply planners validate the lead-time change, and finance presents the scenarios to the leadership team with agent-drafted commentary that analysts have checked. The decision is made in days rather than after a quarter of surprises, and every assumption has an owner.
Common Pitfalls
- Starting with technology instead of drivers. Without agreed drivers and definitions, a connected platform simply connects inconsistent numbers faster.
- Over-centralizing. If finance tries to own every operational detail, functions disengage and maintain shadow plans.
- Linking at the wrong granularity. Connecting SKU-level supply plans directly to the P&L creates noise. Agree on aggregation levels for each link.
- Ignoring data latency. A linked model refreshed monthly delivers little more than a disconnected one. Align refresh cadences with decision cycles.
- Letting agents act without guardrails. Agents should propose changes to plans, not commit them, until controls and audit trails are proven.
- Measuring only cycle time. Faster planning is valuable only if forecast accuracy and decision quality improve as well.
Choosing Tools: Platform or Composable?
There are two broad architectural paths. A planning platform approach uses a single vendor’s suite, such as Anaplan, Pigment, Workday Adaptive Planning, SAP Analytics Cloud, Oracle or Board, to host most functional plans. It offers integrated modeling and increasingly built-in agents, at the cost of vendor dependence. A composable approach keeps functional plans in their native systems and uses a data platform such as Databricks, Snowflake or Microsoft Fabric as the integration and analytics layer, with planning tools consuming shared drivers from it. Many organizations end up with a hybrid: a primary planning platform for finance and closely linked functions, and the data platform as the source of actuals, operational detail and forecasting models.
When comparing options, ask how each handles shared driver governance, how quickly operational data reaches planning models, what agents can do and under whose permissions, and how scenarios are versioned and audited. Product demos rarely answer these questions; a short proof of concept on one critical link does.
Metrics That Show xP&A Is Working
Because xP&A spans functions, success needs to be measured across them rather than only inside finance. A balanced scorecard keeps the program honest and prevents it from becoming a technology project that reports progress in features delivered.
- Decision latency: the elapsed time from a material change in a driver to a reviewed, cross-functional scenario in front of decision makers.
- Forecast accuracy and bias for revenue, operating expense and cash, tracked by horizon, compared against the pre-xP&A baseline.
- Plan consistency: the number of unresolved differences between functional plans and the financial plan at each cycle.
- Manual effort: analyst hours spent on data collection and reconciliation versus analysis, measured through simple time sampling.
- Agent quality: the share of agent-drafted variance explanations and scenarios accepted with minor or no edits.
Outlook
Over the next two years, expect planning to become more continuous and more automated at the edges, with agents handling monitoring, preparation and first drafts while planners focus on judgment and trade-offs. Zero-copy data sharing should keep reducing integration effort, and semantic layers will increasingly hold the shared driver and metric definitions that both planning tools and agents rely on. The organizations that benefit most will be those that treat xP&A as an operating model, with clear ownership of drivers and links, rather than as a software purchase.
Conclusion
xP&A addresses a long-standing problem: plans that make sense individually and conflict collectively. What is new in 2026 is that the enabling pieces, connected data, embedded AI agents and stronger CFO demand for decision impact, have matured enough to make it practical. Start with one painful link, agree on shared drivers, connect plans at sensible levels of detail, and introduce agents once the foundation is trusted. Done this way, xP&A shortens the path from a change in the business to a decision about it.
How Hanoi Technology Can Help
Hanoi Technology helps finance and operations teams design shared driver models, connect planning data and introduce AI-assisted xP&A in stages.