An ERP program needs about a dozen roles filled on your side of the table, and the ones that decide the outcome are the ones you cannot buy from the SI. The sponsor, a client program manager who is not an SI employee, a business process owner for each core area, a data lead who owns the reconciliation sign-off, and an independent assurance lead who reports to the sponsor. Each role you leave vacant or hand to the SI shows up later as a warning sign the status report hides. Gartner predicts that by 2027 more than 70% of recently implemented ERP initiatives will fall short of their original business case. Most of those programs will go live. The gap is almost always a role nobody owned.
The roles you own, and the ones you can buy
The SI brings consultants, configuration skill and a method. It cannot bring the decisions, because the decisions belong to the people who will run the business on the system for the next fifteen years. The division is simple. The SI supplies hands. You supply judgment, ownership and sign-off. When those move to the SI, the program still runs, but it runs toward the SI's incentives, which are the date and the fee, not your business case.
The table below lists the client-side roles and what each one owns. The last column names the warning sign that appears when the role is vacant or outsourced. The signs and their thresholds sit in the companion piece on the seven signs a program is in trouble, and are not repeated here. To put a name against each role before you sign the SI contract, work through the roles-to-fill checklist.
| Role | What they own | If vacant or handed to the SI, it surfaces as |
|---|---|---|
| Executive sponsor | The business case, and every decision above the program manager | A steering committee that meets and decides nothing |
| Steering committee | Scope, date and budget trade-offs, in writing | Scope sliding to a phase 2 with no budget line |
| Client program manager | The plan, the SI, and the status the board sees | A RAG status that stays green while the records turn |
| Business process owner, per core area | The design decisions the system will enforce | Signed designs reopened after build starts |
| Global process owner, multi-site | One design that holds across sites | Local variants that cost rework and a shared template |
| Data migration lead | Cleanup and the reconciliation sign-off | Mock loads that slip or shrink before cutover |
| Integration owner | Every interface and its fallback | Deferred integrations that become manual workarounds |
| Security and controls owner | Segregation of duties and the approver gate | An audit finding at the first close |
| Change and training lead | Adoption after go-live | Operational disruption in the first weeks live |
| Super-users | Local knowledge and day-one support | A help desk that cannot cope and workarounds that stick |
| Independent assurance lead | An honest read of the records, for the sponsor | Drift caught in the week before cutover, not two quarters out |
The sponsor decides, and the steering committee decides with them
An executive sponsor who cannot move money, change scope or move the date is a figurehead, and the program will find that out at the first hard decision. The sponsor owns the business case, which means the sponsor is the person who loses if the program goes live on a case it can no longer meet. That is why the sponsor, not the program manager, must chair the steering committee and sign the scope cuts.
The steering committee has to decide things. A program spending $1M a month generates at least one decision a month that belongs above the program manager: a scope cut, a date move, a draw on contingency. If the committee meets for 40 minutes and records no decision, those calls are being made one level down, by people with less authority and a stake in the date. We recommend monthly meetings in the year before go-live, each with a written decision log, and sponsors who attend rather than send delegates. When the meetings get shorter and the log stays empty, the committee has stopped governing, and that is one of the warning signs.
The client program manager cannot be an SI employee
The person who writes the status the board reads must work for you. A RAG status is an opinion about a plan, and when the SI both runs the plan and grades it, the opinion rounds up. By the time the color changes, the defect log and the data loads underneath it have usually been moving the wrong way for a quarter or two. A client program manager reads those records and reports what they show, because that person's incentive is a working system, not a hit date.
This is the role buyers most often concede, because the SI offers to supply it and it looks like a saving. It is the most expensive concession on the list. An external program management office runs $5,000 to $20,000 a month by ERP Research's figures, and a client program manager costs the same order. Suppose a $300M manufacturer lets the SI provide the program manager to save a year of that cost. The manager reports green for three quarters while open defects climb, because the manager's firm is paid to hit the date. The saving buys a late surprise, and a rescue costs 30 to 50% of the original budget by Deliberate Directions' NetSuite guidance, a single advisory firm's figure best treated as a planning range.
Business process owners own the design, and a global owner holds it across sites
A business process owner for each core area, order-to-cash, procure-to-pay, record-to-report, plan-to-make, is the person who decides how the business will run on the system and signs the design. Not the IT analyst, and not the SI consultant. When a design document carries an IT signature and no process owner's name, the sign-off was procedural, and the owner sees the design for the first time in testing. That is when signed decisions get reopened, and anything reopened after build starts is a change request the SI is right to charge for.
The failure mode is well understood. Owner attendance below 75% at design and test sessions, or an owner with less than half their time freed, means the day job wins every scheduling conflict and the decisions get made without them. The fix is backfill, covered below, and it is the cheapest intervention on the list.
Where one design must hold across several plants or countries, name a global process owner who outranks the local owners for that process. Without one, each site argues for its own way of working, and you either pay for the variants or relitigate them after go-live. Each additional legal entity can add 10 to 20% to the implementation, by CRM Layers' estimate, a weak figure best read as a direction rather than a precise cost. The global owner is how you hold that number down, because a single agreed design costs far less than one negotiated plant by plant.
The data lead owns cleanup and the reconciliation sign-off
Data is the top cause of schedule overrun, which Panorama Consulting's 2025 ERP Report found had replaced resource constraints across 172 organizations. The data migration lead is yours, because the records are yours and the cleanup is work only your people can do. The SI can build the extraction and mapping. It cannot know that an asset is retired, a customer is dead, or a work order's location code is wrong. Only the business does.
The sign-off is the part that must stay on your side. The data lead owns the rule that at least three full-volume mock loads run before cutover, that the second loads 95% of records cleanly, and that the controller signs a trial balance variance between the legacy and loaded ledgers. "Data migration 80% complete" usually means 80% of the mapping objects are built, which is not the same as 80% of records loaded, and neither is a trial balance the controller will sign. When the mock loads slip or quietly shrink in scope, the data lead is the person who catches it while there are still months to run cleanup.
Integration, security and change each need a named owner
Three roles are easy to leave implied and expensive to leave vacant.
The integration owner holds every interface and its fallback. A go-live design that assumes an interface which then slips to phase 2 means a manual workaround at cutover, and workarounds are where operational disruption comes from. The owner's job is to know which integrations the go-live depends on and to refuse to let any of them become a phase 2 item without a decision.
The security and controls owner designs segregation of duties into the system rather than bolting it on after an audit finding. This role now carries a second duty. Where an AI agent or the SI drafts a configuration, a mapping or a posting rule, a named approver on your side has to sign it before it is built. A configuration no one can trace to a named approver is a control gap under SOX, as the piece on governance for AI agents sets out. The controls owner holds that gate.
The change and training lead owns adoption, which is the benefit the business case actually counted. A system that goes live clean and then is not used returns none of its case. This lead runs the training, the super-user network and the first weeks after go-live, and reports adoption to the steering committee as a number, not a reassurance.
Super-users learn the system, and every lead you pull on gets backfill
Super-users are the business people who learn the system deeply enough to design it, test it and support their colleagues on day one. They are the thread running through the whole program, and they only work if they are freed to do it. A super-user still carrying a full day job attends when they can, which is not enough, and the knowledge that should have stayed in the business leaves with the SI.
Backfill is the discipline that makes every other role real. For each business lead and process owner pulled onto the program, backfill their operational job for the design and test windows. A tier 1 program needs 36 months from decision to go-live, which is years of senior people's time that the business still has to run alongside. Testing sign-off belongs to the business, not the SI, and a test signed only by an SI consultant and an IT analyst has not been accepted. The owners can only sign if they were in the room, and they are only in the room if someone else is doing their day job.
Independent assurance reports to the sponsor, not the program
The last role sits outside the program by design. An independent assurance lead reads the source records, the defect log, the mock-load reconciliation, the change register, the steering minutes, and reports to the sponsor rather than to the program manager. The reviewer must not depend on the go-live date for a bonus, which rules out the SI's own quality review, useful as a second line but never the independent one.
Budget 1 to 2% of the program for it, the rule set out in the companion piece on independent program assurance. On a $12M program that is $120,000 to $240,000. It pays for itself because it catches drift about two reporting quarters before the status color changes, while there is still time to act, rather than in the week before cutover when the only choices are a delay or a bad go-live.