A company rolls out Microsoft Copilot in January and trains everyone who gets a license. By April the training is finished, the licenses are live, and most people are working the way they did before. The training was good, but nothing followed it.
Nudges in Activate reinforce new behavior in the weeks after a course ends. Campaigns widens what a sequence can be built around. An administrator can now build one around any business priority, aim it at any team, and launch it at any time, without filing a support ticket. All of it is live in Activate today.
The training itself is rarely the problem. Customers put real work into their programs and get good feedback on them. What goes wrong is the weeks afterward. Ebbinghaus measured the decay more than a century ago, and about 70% of new learning is gone within 24 hours if nothing revisits it. Activate's nudges are built on nudge theory and the forgetting curve for that reason.
Haunstrup and Jensen (2024) tested the fix in a randomized controlled trial of 226 managers and 4,442 employees. Both groups got the same training. The group that also received just-in-time nudges afterward still showed changed behavior eight months later. The training-only group did not.
Most organizations may never even notice this learning loss because they measure the wrong things. Donald Kirkpatrick laid out four levels of training evaluation in 1959: reaction, learning, behavior, and results. As Why Course Completions Don't Prove Your L&D Is Working covers, most organizations still report on the first two.
For starters, people will not keep leaving their work to log into another tool and be reminded of what they learned. And teams start from different places, so the reminder that helps one group is noise for another.
The second problem defeats one-size-fits-all follow-up. Two departments can post the same flat usage number on a new AI assistant for opposite reasons. In one, managers never showed people how the tool fits into their work. In the other, people worry that leaning on it will erode skills they spent years building. EY's Work Reimagined survey of 15,000 employees found 38% fear AI will replace parts of their job and another 38% worry it will erode their skills. In the COM-B model, the first group has an opportunity barrier and the second a motivation barrier. The same reminder will not fix both, which is the gap Campaigns exists to close.
A campaign is a set of nudges built for one audience and one goal, scheduled by an administrator, and delivered in the tools people already use. A priority named on Monday can be reinforced in the field that same week. Nudges come from two places. Perceptyx's library holds roughly 3,500, each built on nudge theory and mapped to one of the 90 behaviors in the People Insights Model, delivered by the Nudge Agent through Microsoft Teams, Slack, email, and mobile tools for frontline employees. The Nudge Generator builds custom nudges from the company's own content, and both kinds can run in the same campaign.
AI-Assisted Action Planning is still its own capability inside Activate. It starts from survey results and helps managers pick actions for their teams. A campaign starts from a business goal, so it does not wait for a survey and does not run through a manager's action plan.
Competitors will say any platform can schedule messages, which is true but misses the difference. Scheduled messages tell someone to do something. A nudge tells them what to do and how to think about it, why it matters for them, and how to start in the next few minutes. They can open a coaching conversation in the same thread if they want help. Because each nudge maps to a behavior, a leader can check later whether the behaviors the campaign targeted are showing up.
Audiences are built with filters for team, department, function, and level. A campaign can reach the claims department, or every manager in one region, and nobody else, so one group's priority never spends everyone else's attention. When Perceptyx wrote about building custom nudges in June, targeting worked by leadership level rather than by named group. Campaigns narrows it to the group the goal is for.
Two conditions still decide who gets a given nudge: whether the person is in the audience and whether they hold the relevant behavioral anchors. Custom nudges are prioritized in the queue, so a qualifying person's next nudge is likely to be one of them. A nudge that matches what someone is already working on gets acted on, and that relevance keeps people opening nudges.
Start with the goal rather than the content. The more specifically an administrator names the behavior and why it matters, the sharper the nudges that the Nudge Generator returns. Next, upload one piece of source content: an article, a research report, the company's AI usage guidelines, an existing training deck. The Nudge Generator builds nudges from it, so the campaign carries the company's own guidance in the company's own voice. When Perceptyx ran this internally, three custom nudges arrived in under two minutes.
Most of the administrator's time goes to editing. Each nudge's wording and behavioral anchors can be changed, a Custom Link can point to a related resource, and the nudge can be translated. Every custom nudge carries a "Why this nudge?" line naming its source and focus area, so the person receiving it knows why it arrived. Then publish, set the audience, and schedule. The whole thing takes minutes.
Campaigns serve goals such as AI adoption at scale, culture change, manager and leadership effectiveness, and learning reinforcement. The example below describes the first of these.
A specialty insurer bought Microsoft Copilot licenses for claims and underwriting in January and published AI usage guidelines alongside them. By April, 71% of licensed employees had logged in at least once. Two claims teams used it daily. Everyone else had tried it twice and gone back to the old workflow. The company had paid for a year of licenses and a full training rollout, yet neither department was closing work any faster than it had in December — no efficiency or productivity improvements had occurred.
Why had adoption stalled? The HR director who administers Activate put that question to managers in both departments and got two different answers highlighting two different behavioral problems. Claims adjusters liked the tool but had never been shown where it fit into a claim, so the opportunity to use it never appeared in their work. Underwriters had enthusiastic managers but worried Copilot would dull judgment built over fifteen years, which is a motivation barrier. Both groups produced the same login number and needed opposite responses.
She built two campaigns that afternoon. The first went to claims managers, filtered by department and level. She uploaded the company's AI usage guidelines to the Nudge Generator, edited what it produced, and added library nudges on coaching conversations. One asked each manager to pick a recurring task in their next one-on-one and work through it with Copilot alongside the adjuster.
The second went to underwriters. Those nudges asked them to use Copilot on one file a week and note what they accepted and what they overrode. She wanted underwriters governing the tool with their own judgment, which is a different behavior from using it more.
Both campaigns ran for eight weeks in Teams. Before they launched, she added four AI readiness questions to the June employee pulse survey so she would have something to measure against. In June she brought three numbers to the executive review: readiness by function, the share of each group that acted on the nudges, and cycle time on the claim type she chose in January.
She could make that case because she picked the metric and took a baseline before the campaign ran. A campaign can't prove a shift in perspectives or behaviors on its own. Without a baseline, there is nothing to compare against afterward. Activate tracks nudge engagement and movement on the behaviors a campaign targets, in survey items or in the work itself, and with Develop it tracks learning retention too.
Among employees receiving nudges, 96% report a noticeable positive change in the behaviors the nudges reinforce. Customers using nudges see up to a 12-point engagement lift and 20% improvement in targeted behaviors within six months. Perceptyx's own custom nudges in June drew reaction rates 11 and 20 points above baseline, measured across one internal population in the first weeks after launch.
Develop turns a company's existing training content into a conversation with each learner and checks whether they understood it during the lesson. So by the end, the company knows which concepts each person struggled with. A campaign can then send nudges on those exact concepts over the following weeks, while people are working. For L&D teams, the report to leadership changes with it: instead of completion and attendance counts, they can show who gained a capability and whether it is showing up in the work.
Nudge Boost and Org Behaviors now run inside Campaigns. The targeting and reinforcement customers have relied on are still there, and administrators can now create custom nudges, set audience filters, and schedule campaigns without a support ticket. Perceptyx's Workforce Transformation Consultants will move each customer's existing Org Behaviors into Campaigns and can set up the first campaigns alongside any admin team that wants the help. Easier-to-read nudge engagement metrics and downloadable reports are next on the roadmap.
Pick one goal for this quarter, one audience, and one number you can re-measure in 90 days. Then schedule a meeting with a Perceptyx expert to build the first campaign with your own content and channels. To see how Campaigns fits with the Nudge Agent, the Coaching Agent, and the rest of Activate, visit the Activate page. For the research on reinforcement in the flow of work, read the Beyond Course Completions white paper.