AI tools are now part of most project delivery processes, and that creates governance challenges for project managers. With 70% of PMs working in organizations that use AI and 29% in organizations that plan to use AI, someone needs to own the AI-produced outputs to protect project outcomes. PMs need to develop processes to review, contextualize, and refine those outputs, and not just at the end stage of the workflow.
Because AI assists with so many project inputs, synthesis and human discernment need to happen at each output stage, to make sure that the project proceeds in the right direction. Project managers, team leads, and other team members using AI tools need a playbook that ensures there’s a person accountable for each AI result that’s used to make decisions or take action. This calls for a mindset shift in how project managers think about AI-related project work.
What does AI output synthesis look like in project management?
AI output synthesis creates a layer of human context and judgment between what AI tools produce and what the team uses. For example, when an AI image generation tool creates a set of marketing images, a team member who knows the brand guidelines and the overall context the brand operates in needs to check them. They’ll look for alignment with core branding requirements like color palette, fonts, and logos. They’ll also check the current events context, to make sure the images don’t call to mind any unwanted associations, like running aviation-related imagery in the aftermath of an air disaster.
UI/UX teams need the same control layer to make sure their designs line up with the way their actual customers will use the site, form, or app. The same information may need to be expressed differently for different audiences. If an AI tool recommends “tuition and fees” as part of a college’s site navigation copy, it will be up to someone on the team to decide if that phrasing works or if “paying for school” is a better choice for their audience. The same principle applies across any project with AI inputs, and it creates work that project managers need to allow for in their timelines. While AI and agents may enhance, accelerate, or even replace certain tasks, the time and effort required to understand, digest, and educate stakeholders on the outputs is critically important.
The evolution of context matters, too. An AI-generated contract based on an approved template may have all the required elements, but it may not be fit for purpose without human review. For example, a templated contract for a renewing client may not account for changes to risks, assumptions, and timelines that have happened since the previous contract was signed. A team member or lead needs to adjust the language to factor in new priorities and dynamics before sending it to legal for approval.
The time factors that AI output synthesis impacts
As AI accelerates delivery velocity, it’s easy to assume it also compresses timelines. AI acceleration is also changing the time required for, and the timing of, output synthesis. Work planning is a useful example. When a PM feeds past project data into a new project planning tool, the AI will make planning assumptions based on that data. So, if the AI tool only receives data from successful projects that met deadlines, it will use that data to generate optimistic scenarios and timelines. These outputs can be a useful starting point, but they need to be evaluated in the context of current conditions. If staffing levels, budget, or other factors have changed since the past projects, those plans will need to be adapted.
At the same time, output synthesis needs to happen at every project stage where AI is involved. That takes time that project managers must factor into their planning, so their teams can benefit from AI acceleration without getting bogged down by the need to correct AI-related issues later in the project.
The skills PMs and teams need to synthesize AI outputs
The central skill required to assess and adapt AI outputs is discernment. Stakeholders responsible for AI output synthesis at any stage of a project should know if an output will be useful, given what they know about the project, the client, customers, and other factors. These team members and leaders also need to feel comfortable making judgment calls on outputs, based on business goals and ethical standards, as part of their accountability.
Taking a step back from individual outputs for projects, PMs and other decision makers also need to know which AI tools are right for specific tasks and which data those tools can use. This matters for compliance and customer trust as well as project outcomes. For example, some tools are meant for secure internal use with sensitive data, but others are not. Using approved tools and avoiding the use of shadow IT by team members is another element of discernment. So is knowing which AI-provided results to use when different models return conflicting results.
Critical thinking and good judgment matter for AI inputs, too. Just as giving an AI tool only data about successful projects will bias it in favor of optimism, each input needs to be structured with the right reference data, prompts, and contextual information to deliver useful results based on realistic parameters. Using discernment to get reliable results from AI tools is an important skill that develops through regular use, including a willingness to experiment and the time to do so.
Developing AI output synthesis best practices
Project managers are the ideal people to build processes for governing AI outputs, because they already know how to read a room, evaluate information for context, communicate with different audiences, and allocate time to get work done. Synthesizing AI outputs is a new way to apply those skills. PMs who build these ongoing review steps into their projects from the beginning will be able to move faster and with more confidence in their final product than teams that treat AI output review as a final-stage step.
