AWS Deal modeling

Streamlining complex workflows

The opportunity

In 2023, the AWS Fintech team launched ADAPT to streamline deal modeling between AWS service finance teams and GDSP/BD Financing. Despite targeting the four largest services, the rollout fell short: EC2 declined to onboard due to unmet critical requirements, and adoption among other services was just 17%. To address this, we focused on uncovering adoption barriers and aligning the product with EC2's mission-critical needs.

my role & the team

user research

Prior to my joining, the team had conducted user research for the initial launch. Using that research as our starting point I decided we should conduct some tasked based shadowing sessions to identify existing pain points and obstacles for our users.

For our EC2 partners I scheduled interviews to better understand the business requirements that were not accounted for in the initial launch.

We shadowed 3-5 analysts from each of our 4 Phase 1 service teams as well as the core GDSP/BD finance team.

We also conducted 3 interview sessions with EC2 Service Finance Managers and Analysts to ensure we were addressing their business critical concerns

Based on the insights generated from our research we categorized the work needed and began prioritizing with our cross-functional teammates. 

goals

We established key goals and milestones to define project success, divided into two phases:

Phase 1:

• Onboard EC2 Service Finance team to ADAPT and boost adoption from 17% to 70%.

• Reduce deal modeling time by over 70% (from 2–4 hours to under 1 hour).

• Cut deal approval time by 50% (from 10 to 5 business days).

Phase 2:

• Assess operational cost savings.

• Evaluate deal accuracy (utilization, margins, escalations, renewal rates) over its lifecycle.

concept testing

Collaborating closely with Product, Engineering, and our users, we embarked on redesigning the ADAPT experience. To maintain alignment, I introduced monthly user testing sessions. During these sessions, we shadowed users as they completed task-based activities using clickable prototypes. We measured key metrics, including time to completion, number of clicks, error rates, and overall system usability, ensuring continuous feedback and iterative improvements.

final designs

impact

We launched the updated version of ADAPT in last October 2024. We monitored performance against Phase 01 goals for 6 weeks. Below are the outcomes to goals.

• 382% increase in adoption (from 17% to 82&%)

• 84% reduction in time to deal modeling completion and submission (3 hrs down to 30 mins per deal)

• 91% reduction in clicks to complete a deal submission (178 clicks to 16 clicks)

• 40% reduction in days to approval (10 days to 6 days)

Prior to the redesign each of the 4 services the Phase 01 roll out supported had a very customized workflow. This initial solution was not scalable nor sustainable as we moved to onboard the remaining 17 services with AWS. As we embarked on the redesign I stressed to the team that we needed to design a templated workflow, allowing for some customization, within the templates themselves, to meet the needs of the different services offered. This would now allow for easy maintenance, a unified experience, and limit scope for onboarding new services.

next steps

For next steps in 2025, we will continue to monitor of Phase 01 goals to assure that we remain on or above target. Specifically we will monitor the time to approval since we were unable to meet our 50% reduction in the initial relaunch, attributed to some P2 features we deprioritized to 2025. Below are our other next steps for 2025.

• Prioritize and launch P2 features

• Monitor Phase 02 goals

• Begin work on integration with our P&L reporting platform, PRISM

• Launch Customer Central 

learnings

One of our key learnings in developing ADAPT centered around refining our research methodologies. Several missteps impacted our ability to design a product that truly met user needs when initially launched in 2023.

Key Learnings:

• Stakeholder focus mismatch: The research heavily involved senior leadership, who were not the primary users of ADAPT. Greater emphasis should have been placed on Finance Managers and Analysts.
  

• Limited understanding of workflows: User interviews prioritized problem-solving and solutioning rather than shadowing frequent users to uncover existing workflows, pain points, and opportunities for efficiency.
  

• Missed follow-up validation: There were no shadowing sessions with core users during prototype testing to validate usability, intuitiveness, and whether the solutions met all necessary business requirements.

customer central

During our research, a recurring theme emerged across all service teams: the need for real-time tracking of client metrics like utilization and margins against deal models. This would shift teams from a reactive approach—relying on deal retrospectives near renewal dates—to a proactive strategy for addressing concerns early.

Meanwhile, as we prepared to launch our AI assistant, Jarvis, we saw an opportunity to elevate its learning models by incorporating deal performance metrics. With this data, Jarvis could predict underperforming deals, enabling smarter interventions.

To explore these ideas, we hosted brainstorming sessions with cross-functional teams to uncover user needs and define desired outcomes. By leveraging established patterns from other products and the insights gained, we rapidly designed Customer Central.

Currently in the shadowing phase, we’re observing users interact with the prototype to identify usability and workflow refinements. Customer Central is on track for launch in Q3 2025, poised to transform how service teams monitor and optimize client performance.