Portfolio, program, contract, financial, organizational, resource, rates, cost, project, earned value, scheduling, supply chain. Every planning discipline in the enterprise is about to share one dataset and one user experience, with best-of-breed functionality tailored per function and orchestrated by agents.
Most large enterprises run multiple distinct planning disciplines in parallel, and those disciplines rarely plan together in any meaningful way.
The program team plans scope and milestones in one tool. Contracts tracks deliverables and modifications in another. Finance owns the P&L and cash forecast in a third. HR runs workforce planning. Delivery runs resource planning. Rates live in a spreadsheet that only a handful of people fully understand. Cost planning sits in the ERP. Project and earned value run in a separate PM tool. Master scheduling and production scheduling sit in yet another tool, sometimes several. Supply chain has its own stack entirely. Portfolio management sits above all of it, trying to make the pieces add up.
Our experience is that the result is the same almost everywhere. The company spends a fortune on planning and still cannot get a current, defensible answer to a fairly basic question, which is what is our plan and is it still true.
Ask who owns "the plan" in any large enterprise and several people, all partially right, will raise their hand. The disciplines below are the ones that show up in almost every organization we work with. Some carry different names from company to company. Some get bundled together or split apart depending on the operating model. The point is not the exact list. It is that every one of these disciplines is a function of the others, and almost none of them are wired that way today.
The top-level view of where the company is investing its money, people, and attention. Strategic mix, prioritization, gating, capital allocation across programs and products, and the trade-off conversations every other planning discipline ultimately rolls up to.
Multi-year scope, milestones, dependencies, technical baselines, and the program-level resource and budget envelope that ties a portfolio decision to actual work in the field.
CLINs, options, mods, deliverables, funding ceilings, and the obligation-versus-funding curve that governs how much work can actually be performed in any given period.
Annual operating plan, rolling forecast, cash, capex, and the consolidated picture finance is held accountable for at the board level. The output most leaders see, and the one most disconnected from the planning beneath it.
Headcount, position management, hiring plans, organizational design, span of control, and the structure that funds every other plan in this list.
Named-resource and role-based assignment to demand, skills matching, utilization targets, and the bench-versus-pipeline view that tells you whether the work can actually get done.
Direct rates, indirect pool composition, wrap rates, forward pricing, and billing rates. The structures that quietly determine whether a contract is profitable or whether the company is paying to do the work.
Cost-element-level budget, allocations, overhead absorption, should-cost models, and the cost structure that prices and reports every job in the company.
Work breakdown structures, networks, time-phased budgets, baselines, earned value, ETC. The execution-level plan that ties effort to outcome and exposes the variance everyone else then has to explain.
Production schedules on the shop floor and the Integrated Master Schedule across program lines. Sequence, capacity, dependencies, and time-phased execution. The discipline that turns a plan into a calendar of actual activity, and the one that exposes whether the rest of the plan is even physically achievable.
Demand signals, MRP, procurement plans, supplier capacity, inventory positioning, and the material plan that ultimately determines whether a delivery commitment is real.
If the portfolio plan, the program plan, the financial plan, and the supply chain plan all disagree about what the company is doing this quarter, the company does not have multiple plans. It has a reconciliation problem it has quietly decided to live with.
Every dimension on the list above has a real owner, a real cadence, and a real reason to exist. None of them stand alone. Change a variable in one and several others have to move with it. Treating these disciplines as independent is, mathematically, wrong, and the cost of that wrong assumption shows up in every cycle.
Watch what happens when a single event hits the enterprise. A contract option gets exercised. That one fact moves through every plan at once.
| Trigger | Dimension Affected | Plan Element That Has To Change |
|---|---|---|
| Contract option exercised | Portfolio | Investment mix shifts, prioritization reweighed, capital allocation revisited |
| Contract | Funding ceiling, period of performance, CLIN baseline | |
| Program | Scope envelope, milestone schedule, risk register | |
| Project & EV | WBS extension, time-phased budget, new control accounts | |
| Scheduling | IMS dates extended, production sequence reflowed, capacity loading and schedule risk reassessed | |
| Resource | Named-resource demand, hiring or contractor backfill | |
| Organizational | Position requisitions, span of control, manager load | |
| Rates | Forward pricing inputs, indirect base, wrap rate impact | |
| Cost | Cost-element forecast, allocation drivers, should-cost | |
| Financial | Revenue forecast, cash curve, margin and EBITDA | |
| Supply Chain | Material demand, supplier capacity, long-lead procurement |
Our experience is that the ripple usually takes weeks. Sometimes it takes months. Often it never fully propagates, and the inconsistency sits in the system until a variance report exposes the gap, almost always at the wrong moment.
This is not really about elegance or good architecture. The practical question is whether the enterprise can see itself in current time. Most of the ones we work with cannot, and the gap shows up wherever a leader has to make a decision faster than the planning system can refresh.
Technology is part of this story, but only part of it. If unification were a pure tooling exercise it would have been solved a decade ago. The harder issues sit across four dimensions at once, and in our experience almost no transformation program addresses all four together.
Planning ownership is fractured by design. Portfolio leaders, program managers, contracts officers, FP&A, HR, delivery leadership, pricing analysts, controllers, project controls, and supply chain planners all report into different leaders, with different incentives, and different definitions of what "good" looks like.
Planning cadences were built around the limits of manual aggregation. Monthly close, quarterly forecast, annual operating plan. Those cycles exist because rolling up the data used to take that long. The constraint that created them is mostly gone now, but the cycles stayed.
The market responded to the planning problem with point solutions. EPM for finance, PPM for projects, HCM for workforce, S&OP for supply chain, rate engines for pricing. Each of those tools is genuinely good at its slice. None of them were ever designed to share a dataset with the others.
Even when the systems talk, the data does not. The definition of a project, a cost element, a position, a contract line, and a resource drifts across tools. Master data sits in several places. Hierarchies diverge over time. Time grain differs. Currency conversion happens three times in three different ways.
One reasonable pushback on unified planning is that pulling sensitive plan data into a single platform increases risk. The concern is fair. A poorly built unified platform really would widen the blast radius of every misconfigured role and every overshared dashboard. A well built one reduces risk, because consolidating dozens of inconsistently governed tools into one consistently governed environment is, by any audit standard, an upgrade. The trade-off comes down entirely to how the platform is built and governed.
In XPM, security is not a wrapper layered on top. It is built into the data model, the agent population, the user experience, and the identity layer at the same time.
Planning data is inherently multi-dimensional. Access cannot be a coarse "this user sees this report." It has to be defined at the intersection of program, contract, organization, cost element, and time, with the policy attached to the data itself rather than the application.
For A&D and regulated industries, plan data routinely crosses CUI and export-controlled boundaries. The platform has to enforce data tagging, partition by classification, and prevent agents from seeing or summarizing data the requesting user cannot see directly.
Every AI agent operates under a principal identity with a defined scope. Agents cannot escalate, cannot read across tenants, and cannot bypass approval thresholds. Every agent action gets logged with full lineage and remains reversible.
Every plan figure, agent suggestion, and user adjustment carries lineage back to the source data and the rule that derived it. Auditors, compliance officers, and the DCAA can answer "how did this number arrive at this value" in seconds, not weeks.
Sensitive plan data does not leave the trust boundary. Agents run against private model endpoints, retrieval is scoped per user, and prompts and responses are logged for review. Customer plan data is not used to train shared models unless that is an explicit decision.
Authentication is continuous. Authorization is contextual. The platform assumes any session might be compromised and limits the maximum damage any single principal, human or agent, can do in any single action.
Every planning vendor for the last twenty years has claimed to deliver unification. Each generation of tool delivered real value inside its own lane, and each one ended up reinforcing the silos around it. Worth understanding why, because XPM is a genuinely different architecture from what came before. It is not a marketing repaint of any of these.
| Era | The Promise | What Actually Happened |
|---|---|---|
| ERP, 2000s | One source of truth for finance, supply chain, and HR | Transactional excellence. Planning kept happening in spreadsheets and side systems. |
| EPM, 2010s | Connected planning across finance and operations | Finance got cohesion. Program, project, and supply chain stayed disconnected. |
| PPM, 2010s | Project and resource planning at portfolio scale | Strong for delivery. Weak on finance integration, missing contract and rates entirely. |
| Best-of-breed, 2015 to present | Pick the best tool for each function and integrate after the fact | Best-of-breed UX, but on disconnected data. Integration cost compounded each year. |
| Data warehouse / lakehouse, 2020s | One place where all the planning data converges | Reporting got unified. The act of planning, of changing the plan, stayed fragmented. |
| XPM / XP&A, now | Best-of-breed UX per function, on one unified dataset, orchestrated by agents | The first architecture where the planning act itself, not only the reporting on it, is unified, and where tailored function-specific experience is a feature rather than a contradiction. |
The pattern across all of those eras is consistent. Each one unified one slice and assumed the rest would catch up. The rest never did. XPM works because the order is reversed. The unified dataset gets built first, planning happens directly on top of it, and every functional view, workflow, and agent inherits from that one model.
The thing the prior eras got wrong was not best-of-breed itself. Best-of-breed user experience is what planners actually want, because the way a contracts officer thinks about CLINs has very little in common with the way a supply chain planner thinks about MRP. The mistake was assuming that best-of-breed UX required best-of-breed data. With modern data platforms and agentic AI, those two things are now separable. You can have the tailored function-specific experience and the unified data spine at the same time, and that is the actual breakthrough.
The technology to build real XPM has only become commercially viable in the last six months or so. Two shifts made the difference. Modern data platforms closed the gap between operational and analytical data, so the same store can carry actuals, plans, and scenarios. Agentic AI made it affordable to apply expert-level reasoning to every plan element continuously, rather than only at quarter-end. Put those two together and the planning platform itself takes a new shape.
A best-of-breed user experience for every planning function. Contracts feels like the contracts tool the team wants. Rates feels like the pricing tool the analysts trust. Project controls feels like the EV cockpit the controls team has been asking for. The UX adapts to the function. The dataset underneath does not change.
Specialized agents per planning dimension, plus orchestrator agents that handle cross-dimension impacts. Agents propose, humans approve. Every action gets logged, scoped, and made reversible.
The rules that turn raw plans into derived plans. Allocations, rate calculations, earned value formulas, MRP logic, headcount-to-cost translation. Logic is declarative, versioned, and testable, not buried inside a tool.
One definition of every planning concept. Portfolio, program, contract, position, resource, cost element, rate, project, WBS, schedule activity, material, and time. Hierarchies, attributes, and grains governed once and reused everywhere.
One physical dataset holding actuals, baselines, forecasts, scenarios, and versions side by side. Every plan, every dimension, every change, in one store with full history and bitemporal awareness.
Contract mod, demand shift, attrition spike, or supplier delay enters the dataset.
Agents propagate the signal across every affected dimension, surfacing what shifts.
Scenario agents generate ranked alternatives with cost, risk, and feasibility scored.
Owners review the trade-offs in a shared view, ask clarifying questions, and commit.
One commit updates every affected dimension at once, fully audited and reversible.
Underneath all of this is what we call solution orchestration. The agents, the planning logic, the semantic model, the tailored function experiences, and the unified dataset are each real on their own. The value comes from how they coordinate. Solution orchestration is the layer that keeps every component working together, so the planner using the tool never has to think about which piece is doing what behind the scenes. It is the discipline that turns a collection of capable components into a working platform, and it is the capability most enterprises will have to build, or partner for, before they can run XPM at scale.
With the right solution orchestration in place, the forecast cycle, the variance review, the rolling outlook, and the program scenario review stop being meetings. They become continuous background processes that people steer when the situation calls for it, instead of executing on a calendar.
Nobody jumps from siloed planning to agentic XPM in one move. There are four stages in the transition, and most enterprises today sit somewhere between the first and the second. Getting to the third is real work, but not insurmountable, and the work gets easier each year as the underlying platforms mature.
Each planning discipline runs in its own tool with its own data model. Reconciliation is manual and monthly. Cross-functional questions take days to answer.
Where most arePlans get integrated via pipelines and a shared warehouse. Reporting is unified. The act of planning still happens in separate tools with separate workflows.
MaturingOne planning dataset and semantic model. All plans get authored against the same store. Workflows are cross-functional. AI assists individual planners directly.
Emerging nowAgents handle continuous propagation, scenario generation, and variance detection. Humans steer rather than execute. Planning cycles dissolve into a continuous flow.
The horizonIn our experience the technology is the smaller of the two problems. The harder one is what happens in the rooms where planning currently lives. XPM only works when the organization is willing to make a handful of consequential cultural shifts.
No single function owns the plan in XPM. Functions own the questions they answer against a shared plan. That shift is uncomfortable for leaders whose authority is currently tied to a specific tool, a specific spreadsheet, or a specific cycle. It also frees them up to do higher-leverage work.
Almost every function today defends its own numbers, because those numbers are the ones it actually controls. In XPM, accuracy lives in a shared, fully traceable source that every function inherits from and contributes back to. Defending becomes unnecessary. That sounds simple on paper. In practice it is one of the harder cultural shifts to make happen, because it asks senior people to trust a source they did not exclusively author.
Manually reconciling spreadsheets has been the badge of honor for a generation of analysts. In XPM that work moves to the platform and the agents. Reconciliation itself does not go away. It runs continuously, in the background, with full lineage. The skill for the humans shifts to provenance, knowing where every number came from, what rule produced it, and what would have to move to change it. That is a more valuable skill and it scales.
Agents handle the propagation work. Humans get to focus on the decisions agents cannot and should not make. Trade-offs across strategy and risk, ethics, customer relationships, and the political dimensions of running a complex enterprise.
Companies that get these shifts right will compress their planning cycles by an order of magnitude and free up their best analysts for the work that actually requires their judgment. Reconciliation does not disappear in XPM. It gets automated. The agents do it continuously against a shared dataset, with full lineage, and the humans operate one level up. Companies that resist the shift will keep doing manually what the rest of the market has handed to agents, and the cost of that gap compounds quickly.
The architecture described above is not theoretical. The pieces you need to build real XPM, a unified data store, a governed semantic model, declarative planning logic, a working set of specialized agents, and tailored per-function user experiences, are all commercially mature today. The work is in assembling them with the right opinions, the right security posture, and a real understanding of how planners actually do their jobs.
Our team has spent two decades inside the planning systems this paper describes as the past. The EPMs, the PPMs, the ERPs, and the rate engines. We have seen first-hand where each falls short and what would be required to replace them with something that actually unifies, without losing the best-of-breed feel that planners count on. Mach12.ai is the result of taking that experience seriously.
We are intentionally understated about the platform because the results speak for themselves the first time a client sees the unified dataset, the agent population, and the tailored function experiences running against their own data. What we will commit to in writing comes down to this: planning is about to be rethought from the ground up, and the gap between the organizations that move early and the ones that wait is going to be much larger than most leadership teams expect.
If you are reading this from a C-suite seat and your instinct is that the timeline has to be longer than we are suggesting, we would push back gently. The technology timeline is short. The cultural timeline is longer, but mostly for the organizations that wait to start. The ones that begin now will reach Stage 3 inside eighteen months and Stage 4 not long after that.
For two decades the planning landscape has expanded outward. More tools, more dimensions, more integration projects, more reconciliation work. The next decade moves in the other direction. The number of disconnected planning systems collapses. The analyst role shifts from assembler to interpreter. Planning cadence shifts from periodic to continuous. The interface to the plan shifts from a grid to a conversation. And for the first time, the platform itself can answer the question every executive has been asking for years, which is what is our plan and is it still true.
XPM is not really a vendor category. It is the operating model that becomes possible when the unified dataset, the governed semantic layer, the agent population, and the tailored function experiences are designed together rather than purchased separately. The companies that build that model first will plan faster, react sooner, and outperform the ones that do not. That gap will be visible inside two years and decisive inside five.
The transition itself is not really optional. The only real choice is whether the enterprise leads it or gets pulled along behind it. The companies that lead will end up with the tailored planning experience their people have always wanted and the integrated truth their executives have always asked for, finally on the same platform. Mach12.ai is where we are headed, and we would welcome a conversation with leaders who are thinking about this seriously.
Revelation Technologies (RevTech) is a specialized SAP consulting and solution architecture firm focused on Aerospace & Defense, professional services, and complex project-based industries. RevTech combines deep domain expertise in program controls, financial planning, contract management, and ERP architecture with a forward-looking approach to AI-native enterprise platforms. Through Mach12.ai, RevTech is building the next generation of agentic, unified planning solutions described in this paper.