This talk explores the problem that workforce training is of increasing worldwide concern and yet has proportionately small (but growing!) representation within the learning sciences. Many well-meaning companies offer training opportunities to their employees, but when push comes to shove, companies feel pressure to push for short-term productivity over learning even if it may end in higher productivity in the long term. One issue is that in the absence of a clear path towards learning that does not compete with productivity, the reality of what it takes to survive as a company exerts pressures that oppose investment of employee time in learning. We believe the answer is to embed learning opportunities into work. Building on over a decade of AI-enabled collaborative learning experiences in the classroom and online, we are working within a new industry practice of Mob programming to create a paradigm for shared cognition in software development so that it will be possible to adjust the priorities between learning and productivity at different times. In this way, it is possible to offer learning opportunities integrated with work. A longer-term focus is quantifying the tradeoffs between learning and productivity to enable more reasoned decision making about prioritization over time, with the goal of offering a forecasting that enables shifting from a short-term optimization paradigm to a more long-term optimization. This talk reports on work that is currently situated within a multi-national online university course using the industry standard software development platform AWS Cloud9 as well as efforts to begin transitioning to industry.