The initial estimate looked promising: AI agents could free up around four hours a week for each sales representative. Then Boston Scientific’s team sat down with the people actually doing the work.
Customer preparation involved information the proposed agent hadn’t accounted for. Some of the administrative work included in the savings calculation wasn’t happening consistently in the first place.
For Jasmine Jaco, VP, Organization Transformation and HR Strategy at Boston Scientific, the experience demonstrated why understanding work matters before making promises about what AI will improve.
In her conversation with Siobhan Savage on The Work Blueprint, Jasmine explained how she is building AI Work Design into an internal practice: involving employees, testing assumptions and connecting individual workflow changes to decisions about the wider workforce.
Jasmine’s role brings together AI organization transformation, work and role redesign, and the development of HR’s ability to lead that change.
One early lesson was about scope. Trying to address the entire enterprise at once would stretch limited expertise too thin. A meaningful area of work offered somewhere to learn, refine the approach and demonstrate what redesign could accomplish.
Sales provided a revealing example. The team initially examined activities such as preparing for customer calls, conducting them and documenting interactions in the CRM. Based on its early assumptions, it estimated that agents could save each representative around four hours weekly.
But job descriptions and high-level conversations could only take the team so far. They described broad responsibilities without revealing all the information, systems and individual practices involved.
Working with Reejig as part of its broader AI work design effort, Boston Scientific was looking beyond job descriptions to understand how work actually happened. Jasmine’s sales example showed why that detail mattered.
Breaking the work into tasks and subtasks, alongside the people doing it every day, changed the picture.
To prepare for customer calls, for example, sales reps used information scattered across different places, including their own phones. The team hadn’t accounted for all of those sources, so the initial agent couldn’t give reps everything they needed to prepare.
The team also learned that many representatives weren’t consistently recording meetings in the CRM. Its estimate therefore counted time savings against work that, in some cases, wasn’t being done. Requiring that documentation through a new workflow could actually add time.
The discovery made the design problem more specific: equip representatives properly for customer conversations and distinguish improvements to documentation from genuine time savings.
Employees’ knowledge had exposed gaps in both the proposed workflow and its business case. The progress was a clearer understanding of what needed to change and which assumptions could withstand scrutiny.
For Jasmine, understanding work at the task level needs to become a repeatable practice across the organization.
Boston Scientific has designed an AI transformation pod structure that brings together specialists in work and organization design, technology, data and AI enablement, alongside business leaders, subject matter experts and program managers. Together, they shape how work will change and how employees will be supported from the beginning.
Choosing where that team spends its time requires discipline. Jasmine considers how many people do the work, how often, the time it takes, the risks, the implications for employees and the cost of the proposed solution. Reejig’s WorkOS provides them a common view of work across business units, roles and regions. That visibility helps leaders decide where to invest in AI and avoid redesigning the same work multiple times across the business.
To support more teams with limited specialist resources, she describes a tiered approach: hands-on support for the most critical initiatives, coaching for others, and reusable methods, resources and technology that help teams tackle less complex work themselves.
Jasmine is also helping her teams account for work that AI creates: reviewing outputs, applying judgment, correcting problems and remaining accountable for the result.
Those steps need time. Under pressure to demonstrate productivity gains, they can easily be compressed or skipped.
That raises a question about development, too. People need experience to recognize when an output is wrong. Decisions about which tasks to automate therefore need to consider how employees build the expertise required to evaluate the results.
Leaders have a role in making that judgment visible. Jasmine wants them to recognize learning and revision alongside successful outcomes, so people can explain what didn’t work and how they improved it.
Her own approach to building support has evolved similarly. Research and presentations can generate interest, but practical examples make work design tangible.
“Proof points create belief.”
The next priority is understanding what individual changes mean together.
Jasmine describes the shared Work Architecture provided by Reejig as a foundation for examining similar work across business units, roles and geographies. That common view can help teams avoid repeatedly redesigning the same work and begin to understand which roles and skills are changing.
Her ambition extends beyond what one agent means for one sales team. She wants to understand what changes across commercial work collectively mean for the workforce over the next three to five years.
Integrating those insights into broader systems and strategic workforce planning remains work in progress.
Her experience points to a concrete leadership responsibility: involve the people who know the work before approving the business case, then account for the judgment, learning and support the redesigned work will require.
Explore more pioneer stories and practical guides to help you redesign work across people and AI.
Explore AI Work Design resources →