In short
- Write the requirements from the week of the person who updates, not from the steering committee's pack.
- Every field a system already holds is a field nobody has to be trained to type.
- An update from a chat or an email must meet the same permissions and approvals as one made in the platform.
The implementation plan arrives as an appendix to the proposal. It runs to twelve weeks. Three of them are configuration. The rest are a change programme: a change lead, a network of super-users, train-the-trainer sessions, rounds of end-user training, a communications calendar and a support desk for the first month.
The business case sits in the next appendix. Its largest benefit is the time the PMO will stop spending on collecting status, and it assumes that every project manager updates the platform every week from go-live. Nobody at the table can promise that, and nobody is asked to.
A year later the licence report shows most seats assigned and few of them in use. The training was delivered as planned. The benefit was not.
Software onboarding for a programme platform is usually planned as a change management exercise, because that is how enterprise software has been onboarded for years. From the buyer's side of the table, I think that plan is often aimed at the wrong person. A PMO software implementation stands or falls on the people who update the plan, and most of the effort of onboarding a new tool goes on the people who read it.
What goes wrong in a PMO software implementation
The problems are set before the contract is signed, in how the platform is chosen and how the plan is costed.
The demonstration is built for the reader. The buyer is a head of PMO or a transformation director, so the demonstration shows dashboards, roll-ups and the board pack. The screen a project manager uses on a Tuesday afternoon appears for two minutes, if at all. The decision is made on the reporting, and the reporting depends on an experience nobody evaluated.
Training is budgeted as the way adoption happens. When training is the only adoption line in the plan, the risk of adoption moves to the users. If they do not update, the explanation is that they need more training, and a second round is bought. Training can teach a screen. It cannot make a weekly task cheaper.
Integrations are left for a second phase. Connecting the scheduling tool, the work tracker and the finance system takes effort, so it is moved after go-live to protect the date. On go-live day people are asked to type into the platform what another system already holds. The platform's figures start out older than the source, and users learn to trust the source.
Licences are counted as adoption. A seat assigned is reported as a user onboarded. The vendor's renewal, the partner's sign-off and the internal project closure can all be met while the plan is out of date.
The business case carries an assumption nobody owns. The hours saved depend on updates arriving without a chase. That assumption is in nobody's objectives, so when it fails the benefit quietly disappears and the platform takes the blame.
What happens after go-live, from the sponsor's chair, is the subject of A platform nobody opens. This post is about the decisions taken before go-live, which settle most of it.
What good looks like
These are the five things I would ask a buyer to decide before signing. Each changes the cost of the weekly update, which is where adoption is won or lost.
Buy for the person who updates
Write the requirements from the project manager's week, not from the steering committee's pack. The pack matters, but every number in it depends on an update someone has to give. My first question to a buyer is who updates the plan, from where, and how many minutes a week it costs them. If the answer is a login and a form, budget for the chasing as well as the training.
In Lodestar, the Polaris assistant reminds each owner what is due in Microsoft Teams, Slack, WhatsApp or the app's own chat, and the owner answers in plain words.
Ask the vendor to show an update without a login
Put one request into the demonstration script: a project manager gives this week's update without opening the platform. Bring a real site report or a set of minutes, with the names removed, and ask to watch it become a proposed change to the plan. A vendor who can only show the form has told you where your training budget will go.
In Lodestar, an owner can reply to the weekly status email, or send a PDF, Word, Excel or PowerPoint file to the assistant. Either one becomes proposed updates for a person to review.
Connect the systems of record before go-live
Move integrations into the first phase and treat them as the core of onboarding. Every field a system already holds is a field nobody has to be trained to type. Agree which system is the source for each kind of data (the schedule, the tasks, the actuals, the documents) and have the platform read from it before anyone is invited in.
Every field a system already holds is a field nobody has to be trained to type.
In Lodestar, Microsoft Project and Primavera P6 schedules are imported and updated in place on re-import, and the integrations read Jira, Asana, Smartsheet, ERP actuals, SharePoint documents and read-only PostgreSQL, MySQL, SQL Server or Oracle databases. Credentials are stored encrypted, signed webhooks send events out, and the MCP server lets AI assistants read the programme under each person's permissions.
Prove it on one portfolio before the licences scale
Start with one portfolio, its real projects and its real owners, and tie the commercial ramp to it. Agree that licences beyond the first portfolio are bought when its owners are updating on time without a chase, and write that measure into the contract. A buyer should not pay for adoption before it can be seen.
In Lodestar, a pilot runs on the client's own data, and the meeting host shows how many projects have been updated for the period and when each owner last updated.
Keep the governance identical in every channel
An update that arrives through a chat or an email must meet the same rules as one made in the platform: the same permissions, the same approval route, the same audit record. If it does not, the auditors will be right to object, and the convenient channels will be switched off. Ask the vendor to show a governed change, a milestone date or a budget figure, made from a chat and arriving in the approver's queue.
In Lodestar, a change proposed in chat is shown before and after, and nothing changes until the person presses Apply. It is then made with that person's own permissions, and where a workflow governs it, it goes for approval and applies once approved.

Questions for your next review
- In the demonstrations we saw, how did a project manager give a weekly update, and did it need a login?
- Which data in the new platform already exists in another system, and is it connected before go-live or after?
- How much of our adoption budget is training, and how much makes the update itself cheaper?
- What measure will tell us in the third month that people are updating, and is it in the contract?
- Does an update sent from a chat or an email pass the same approval route as one made in the platform?
A new tool is onboarded when the people who update it hardly had to learn it. Buy for them, connect what already exists and keep every control in place, and the change programme becomes a small line in the plan.
