Before you can decide where AI belongs, you have to know which work actually matters. I built the opportunity-scoring approach that helps make that decision defensible.
01 · The problem worth solving
Salesforce's internal P2C transformation had a prioritization problem hiding as a data problem. Teams wanted to invest in automation and AI, but they needed a consistent way to compare very different seller and operator problems.
Without one, prioritization risked being shaped by anecdote, visibility and whichever problem had most recently been raised, rather than comparable evidence of where work was actually underserved.
Across seller and operator roles, which jobs are both critical and underserved, and therefore worth further investment?
The signals behind the score
The score combines how critical a job is with the positive gap between priority and satisfaction. Time-spent per micro-job sits alongside it as a decision signal: it shapes how I read an opportunity, without inflating the score.
02 · Why it mattered
At P2C scale, every misplaced automation bet costs twice: the wasted build, and the unsolved friction that keeps taxing sellers. I wanted a defensible, repeatable way to point investment, including Agentic AI, at work that would actually move the needle.
03 · What I did
01
Built the evidence base
Co-led the design of 24 seller and operator persona profiles, structured around jobs, priorities, satisfaction and collaboration patterns.
02
Designed the prioritization approach
Ran the first opportunity-scoring pilot, combining priority and satisfaction to surface underserved work and identify where AI or automation warranted investigation.
03
Scaled it into a living program
Now lead the recurring multi-wave program across 24+ profiles, with survey motions that keep the evidence current over time.
04
Governed the data foundation
Drove the move to Snowflake as the governed JTBD source of truth, so the evidence stays authoritative rather than fragmenting across parallel copies.
04 · The insight that reframed it
The important distinction wasn't simply "painful" versus "not painful." It was what kind of intervention the evidence justified.
Most jobs were relatively well served; friction was concentrated in a much smaller set of micro-jobs. Within that set, administrative work was disproportionately painful despite being rated lower priority, a strong signal for automation. A different group of high-pain, high-priority seller jobs called for direct product or workflow investment instead.
The opportunity map
Seller jobsAdministrative jobsIllustrative & anonymized · real structure, no internal values
The same friction can imply different interventions. Two clusters sit in low-satisfaction territory, but the evidence points them in different directions: lower-priority administrative work may warrant automation, while high-priority seller work calls for direct product or workflow investment.
How to read it
High friction · lower-priority admin
Automation candidate. Pain is high while strategic priority is lower, making automation worth investigating.
High friction · high-priority selling
Improve / invest. Core work that hurts. Worth direct product or workflow investment.
Low opportunity gap
Leave alone. Already well served. Spending here is diminishing returns.
05 · The decision it enabled
Instead of starting with "where can we add AI?", the program created a more disciplined decision: which intervention does the evidence actually justify?
Investigate automation
High friction, lower-priority, repetitive admin work.
A strong candidate to investigate for automation or Agentic AI. The evidence flags it, the team decides it.
Improve / invest
High-friction, high-priority seller jobs.
Where the work is core and painful. Direct product or workflow investment may be more appropriate than adding AI on top.
Leave alone
Jobs already appropriately served.
Where the opportunity gap is small. Further investment is diminishing returns.
06 · Honest status
Where it stands
Live and multi-wave. The scoring and survey motions run on a recurring basis, with JTBD data maintained in a governed source of truth. Turning the map into funded decisions is an ongoing, cross-functional effort. The program produces the evidence and the recommendation; it doesn't own the downstream decision.
07 · How I'd measure value
The program succeeds when the evidence changes decisions, not when another survey is completed. I'd look for:
Decision adoption
How often JTBD evidence is used in prioritization and investment decisions.
Intervention quality
Whether teams can clearly explain why a job calls for process change, tooling, automation or AI.
Evidence reuse
Whether teams use the shared opportunity map instead of recreating fragmented analyses.
Learning over time
Whether recurring waves show meaningful movement in the jobs teams chose to address.