WhereSparq · 2026
ResultReads news on every client and lead. Emails each rep a daily brief. Live with sales.
Every CS agent gets a strip. None flies uncleared.
This is how I'd run Coupa's Customer Success & Support AI portfolio. Every agent idea sits on one board. It moves left to right only when it earns it.
Run the Entitlement Check strip in Test Lab
Scored
Test Lab
Cleared
Live
Onboard
Support
Renew
VAL value score · DEFL est. deflection · HLTH account health lift
Synthetic board. Use cases come from the job description. Scores are illustrative, not Coupa data.
Test lab · ENT-CHK
Entitlement verification. A customer asks what they're entitled to. The agent checks contract, entitlements and usage, then answers with the source attached.
Definition of done
| Gate | Bar | Run 1 | Run 2 |
|---|---|---|---|
| Resolution accuracy | ≥ 99.0% | · | · |
| Same answer on rerun | ≥ 99.5% | · | · |
| Tool calls per case | ≤ 3 | · | · |
| Cites the entitlement source | 100% | · | · |
| Tone rubric, empathy and clarity | ≥ 4.5 | · | · |
| Customer data masked in output | 100% | · | · |
Not runWaiting on the first run
Accurate, compliant, empathetic. That's the bar in the JD, so that's the gate.
Case sheet
Run 0Cases 0 / 1,200Pass 0Fail 0
Synthetic cases and results. The method is Coupa's own: lock the tools, move business rules into the API, and test at volume before GA. Geert Mouwen ran payment batch agents through an 8 week lab of 1,200 batches and 56K invoices before launch (The Spend Stack). I'd hold CS agents to the same standard.
How I'd run the board at Coupa
The strip is the operating model. Six boxes, left to right. Nothing skips a box, and every box has someone who signs it.
IntakeOne door for all of CS
I'm the AI intake lead for CS, Support, Professional Services, Education and Enablement. Problems come in. Specs go out to AI Platform and AI Engineering.
ScoreValue now vs health later
Every idea is scored on deflection, hours and account health, then sequenced on the customer journey.
LabAccurate · compliant · empathetic
No agent reaches a customer before it clears the definition of done at volume.
ClearCS leadership signs
Clearance is a decision CS owns. The steering committee sees gate results, not a demo.
LiveA pulse, not a quarterly deck
Accuracy, usage and escalations on one dashboard, with warnings before a workflow touches a customer.
ValueThe business case closes the loop
Hours and deflection land in opex. Health lift lands in retention. Both report back against the case that funded the agent.
I'm the AI intake lead for CS, Support, Professional Services, Education and Enablement. Problems come in. Specs go out to AI Platform and AI Engineering.
Every idea is scored on deflection, hours and account health, then sequenced on the customer journey. Onboard, support, renew.
No agent reaches a customer before it clears the definition of done at volume. Locked tools. Rules in the API. Misses stay visible.
Clearance is a decision CS owns, not a deploy. The steering committee sees gate results, not a demo.
Accuracy, usage and escalations on one dashboard, with warnings before a workflow touches a customer relationship.
Hours and deflection land in opex. Health lift lands in retention. Both report back against the case that funded the agent.
Platform call
Build on Navi Agent Studio first. Coupa should be its own best reference customer. Reach for Gemini or Claude on the approved platform when a use case needs more. Coupa's own AI Delivery posting names both, and I'm certified on both.
The first 90 days
One agent live, a scored backlog CS leadership owns, and a champion network in every region. In that order of trust, not that order of time.
DAYS 01–30
Listen
- Sit with CSMs, Support and Designated Support Contacts. Map the journey and the ticket mix.
- Inventory every agent in flight, Navi and otherwise, with its owner and its business case.
- Open one intake path for CS, Professional Services, Education and Enablement.
DAYS 31–60
Score & test
- Score the backlog with CS leadership in the room. Pick two agents to take all the way.
- Write the definition of done with Support and AI Solution Architects. Stand up the test lab.
- Recruit the first Success Power Users. Give them a direct line into the backlog.
DAYS 61–90
Clear & prove
- Clear the first agent to Live with its pulse dashboard in place on day one.
- Report value against the business case to the steering committee, misses included.
- Publish the rolling roadmap. Sequenced on the customer journey, owned by CS.
Strips I've already cleared
Real work, real numbers. I build agents myself, put them in front of the people who use them, and stay on the hook after launch.
BUILT · OPERATED · ROLLED OUT OCT 2026
WhereSparq · 2026
ResultGrades reps line by line. Regression tests on the grader. Adopted by the COO.
WhereSparq · 2026
ResultClaude certifications from 36 to 100 in under four weeks. 8% exam fail rate.
WhereLaunch by NTT Data
ResultLaunches 50 to 55% faster. Fewer escalations. Better retention. 12 months.
WhereHCA via Launch
Result36 sites. 745K monthly users. Intake from 15 minutes to 2.
WhereBeam Dental
ResultClaims decisions from hours to seconds.
Stack
Coupa's AI Delivery posting namesGemini · Vertex AI · Anthropic Claude · MCP · Salesforce
I holdClaude Certified Architect (Professional and Foundations), Developer, Associate · Gemini Enterprise Deployment and Agent Development specialist · AWS AI Business Strategist
A note for CS leadership
Your CSMs approve every customer facing action until the pulse says they can stop.
The roadmap follows your customer journey, so it reads like your plan. Because it is.
Success Power Users come first, in every region, before the first agent ships. Adoption is built by people your teams already trust.
And when an agent misses, you'll see it on the sheet before a customer does.
Phil
The full flight plan
My resume, tailored for this role. Same text as the PDF.
PHILLIP MOREHEAD
AI Product Leader · Agentic Workflows · AI Adoption & Change
Westerville, OH · (513) 467-8812 · phil@philmorehead.com · coupa.philmorehead.com · philmorehead.com
Executive Summary
Product leader with 16 years in product. As Director of Product I ran a 60+ person org and owned the budgets, vendors and success metrics behind every launch. I build AI agents myself, take them to production and operate them after launch. Then I do the harder part. I get the people doing the work to trust them and use them every day.
I see customer success from the delivery side. At Launch by NTT Data I moved about 45 client accounts onto a product operating model in 12 months. Launches came 50 to 55% faster, escalations dropped and retention improved. At Sparq I took the firm from 36 to 100 Claude certifications in under four weeks. I also built the agent the sales team reads every morning.
“A terrific product leader, constantly thinking about continuous improvement, for his products and himself. Excellent executive presence: calm, organized, and professional.”
Gopal Kamat, former Manager, Launch by NTT Data
Core Competencies
Agentic Workflow Design & Orchestration · AI Portfolio Strategy & Business Cases · Use Case Scoring & Roadmapping · Agent Evaluation, Guardrails & Monitoring · AI Adoption, Change Management & Champion Networks · Customer Lifecycle & Retention · Model & Platform Selection · Executive & Steering Committee Communication
Professional Experience
SPARQ · Principal Product Strategist & AI Ambassador
Product strategy and delivery consulting firm. The firm’s AI native builder. I turn internal and client problems into agents people actually use.
- Built and operate Account Flare, an agent that reads public news on every client and lead and emails each rep a daily digest of reasons to call or reasons to worry. Rolled out to the sales team in October 2026.
- Built Pitch Trainer, an AI buyer simulator that grades reps line by line against a shared rubric, with regression tests on the grader. Adopted by the COO and rolled out to the Solutions Consulting team.
- Founded and run Sparq Flight School, the firm’s Claude certification program. Took firm wide certifications from 36 to 100 in under four weeks at an 8% exam fail rate. Built the boot camp and contributed to the exam question pool.
- Drove pricing strategy, SOW structure and client artifacts for a Phase 1 AI engagement with a Fortune 50 retailer. Signed October 2026 at $806K.
- Built Agent Ledger, a tamper evident audit trail of AI agent actions and reasoning with masked records, demonstrated on Account Flare’s live calls.
- Led the build of Single Point of Truth (SPOT), the conversational layer of Sparq OS, the firm’s internal AI platform on GCP and Snowflake with role based access.
- Scored 1,100+ dashboards into a phased migration plan for a Looker to Power BI pursuit, with a working prototype behind it.
- Drove the firm’s Anthropic Delivery Readiness Designation submission for Claude Code and Cowork. Registered the firm in Google Cloud’s Delivery Readiness Portal from zero.
NutrAssist · Co-Founder & Chief Product Officer
Family nutrition and pantry app for web, iOS and Android.
- Own product strategy, UX and go to market. Build the product myself with agentic AI and no standing engineering team.
- In beta with ~250 users at roughly 75% still active. Validating willingness to pay ahead of public launch.
Launch by NTT Data · Director of Product
Led a 60+ person product org serving consumer, fintech and healthcare clients.
- Built NTT Data’s first product org. Moved about 45 client accounts from BAs and project managers to product, outcomes and design in 12 months. Launches came 50 to 55% faster, with fewer escalations and better retention. The model has since scaled to hundreds of PMs.
- Led 60+ product managers, owners and scrum masters through layers of leadership, with 15 direct reports, 10 of them product leaders. Grew headcount 150%. Owned budgets and vendors.
- Shipped a hospital iPad intake app for HCA across 36 sites in the US and UK, reaching 745K monthly active users. Cut patient intake from 15 minutes to 2. Integrated Epic and Meditech data over FHIR and HL7, and owned device rollout, MDM and field reliability. Partnered on HIPAA, GDPR and NHS requirements.
- Took AI to production for a global athletic footwear brand across roughly 200 US stores. A hybrid ML and rules model more than doubled forecast accuracy. Stockouts fell 37% and sales rose 20% on popular drops.
- Set the success metrics before every launch and the now, next, later roadmap standard across the org.
Launch by NTT Data · Consulting Manager
- Owned product strategy for the firm’s largest account ($20M) as executive escalation point. Grew revenue ~$1M by launching new product lines.
Safewire · Director, Product & Operations
- Built the roadmap and go to market for consumer wire fraud prevention in real estate payments, where trust was the product. Scaled technical and production teams 300%.
Beam Dental · Senior Product Manager
- Shipped an internal claims adjudication and workflow tool that cut claims processing from hours to seconds.
- Launched partner onboarding on both sides of a regulated insurance market, bringing new benefits platforms online and adding ~$250K annual revenue.
State of Ohio, Dept. of Developmental Disabilities · Business Process Analyst / Systems Lead
- Led identity and access management (OAuth, SAML) for HIPAA sensitive applications. Cut report generation from 45 minutes to 30 seconds for statewide program staff.
Earlier Experience
- Creative Arts Lifestyle, Senior Product Manager (2016 to 2017): Led an e-commerce replatforming of 50K orders and 25K products.
- Client Solution Innovation, Product Manager (2014 to 2016): Took B2B products from concept to launch. Grew one client’s revenue 417%.
- Better Together, Product Manager (2010 to 2014): Led web and mobile products and a 12 person team. Scaled adoption from 4 to 77 campuses.
Selected Builds
- Model Compare: a blind bake off of four Claude models building the same page. modelcompare.philmorehead.com
- AI Went Wrong: six interactive scenarios on how AI fails and what each failure costs. sparq-ai-went-wrong-scenario-one.netlify.app
Education
B.S., Health Administration · Minor: Business Administration · Ohio University
B.S., Cybersecurity & Information Assurance · Coursework completed · Western Governors University
Certifications
- Anthropic: Claude Certified Architect (Professional, Foundations), Claude Certified Developer (Foundations), Claude Certified Associate (Foundations), Claude Code Partner badge. The only person at Sparq who is four for four.
- Google Cloud: Certified Partner Specialist, Gemini Enterprise Deployment; Certified Partner Specialist, Gemini Enterprise Agent Development.
- AWS: Certified AI Business Strategist. Demonstrated microcredentials in Agentic AI, Securing Agent Identities, Databases for Agentic AI, MLOps and Serverless.
- Scrum.org: PSPO I & II, PSM I & II, PPDV, PAL, PSM AI Essentials, PSPO AI Essentials. SAFe POPM.
Put me on your board. I'll earn the clearance.
Thirty minutes is enough to walk your real backlog through this lifecycle.
Sources for the Coupa facts on this page
- Director, AI Products, Customer Success & Support (Coupa, Lever)
- How we built deterministic Payment Batch Creation Agents (Geert Mouwen, The Spend Stack)
- Sr Director, AI Delivery (Coupa, Lever)
Built by Phil Morehead with Claude. Board data and test results are synthetic and illustrative. Not affiliated with or endorsed by Coupa.