A project can be delivered on time and still struggle. Staff may keep using the old process, the owner may become the point of approval for every decision, or the new system may solve a problem the business never properly defined.
Across startups and established organisations, the useful question is where progress is getting stuck. Three patterns deserve attention: support after delivery, clear responsibility during the work, and the operational foundations underneath the technology.
Plan for the weeks after launch
A working system is only part of the result. People need to understand when to use it, how it fits their responsibilities and where to get help. Without that, a new process can sit alongside the old one and create more work.
Include adoption in the project plan. Set aside time for training, give staff instructions they can refer to, and nominate someone to collect problems from actual use. A short follow-up after launch can reveal gaps that were hard to see in a demonstration.
- Check whether staff are using the system for the tasks it was built to handle.
- Find out why work is still happening in spreadsheets, messages or duplicate records.
- Agree who will make corrections and provide ongoing support.
At BMI, training and support are part of the conversation about delivery. We agree on what is needed for each engagement so the team has a practical way to use the work after handover.
Give decisions a clear owner
Founders bring important knowledge about their customers and their business. A project needs that involvement, alongside enough delegated authority for the delivery team to work.
Write down the decisions the owner needs to make, the decisions the team can make and the point at which a change needs discussion. When priorities shift, record what changed and what it means for the scope, budget or schedule.
This helps avoid two common problems: work waiting for one person’s approval, and work drifting because nobody has made a decision. We work closely with owners on the direction of a project and raise concerns early when an idea needs adjustment.
Fix the process before adding technology
Before commissioning an app or an AI tool, describe the operational problem in ordinary language. Identify who experiences it, how often it happens and what a better result would look like. Then check the data, responsibilities and existing workflow.
RAND’s 2024 research, based on interviews with 65 experienced AI practitioners, identified unclear problems, inadequate data and a focus on technology over user needs among the causes of AI-project failure. Its scope covered AI and machine-learning development, rather than simply using a pretrained language model. Read the RAND report.
For a business, the practical starting point may be modest: organise records, remove a duplicate approval, document a task or train the person responsible. That makes it easier to judge whether software or automation will help.
Review the foundations before expanding
Before adding another feature or increasing the project’s scope, check that the current work is being used and producing the intended result. Keep the owner involved, give the delivery team clear responsibilities and make follow-up part of the plan.
If your project has stalled between those pieces, tell us where it is up to. We can discuss the strategy, systems and support it needs.
References and further reading
Research, commentary and industry examples for further reading. Read findings in the context of their dates, methods and settings.
