Ep200: Scaling and Monetizing AI-Powered Products with ServiceNow, TwelveLabs and AWS

Mar 31, 2026

Guest

Rich Geraffo

Rich Geraffo, AWS leader helping ISVs scale AI and cloud; Krish Ganapathy, ServiceNow AI strategist who built evaluation-first cultures and prompt engineering roles; Jae Lee, TwelveLabs CEO building multimodal video intelligence. They discuss scaling and monetizing agentic AI, creating new org roles and data governance, token-driven adoption and video RAG at scale, and real-world production tradeoffs and trust challenges.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

ADVICE

Create Prompt Engineering Roles And Evaluate First

Create new AI-specific roles and shift to evaluations-first testing to embed AI thinking into engineering culture.
Krish at ServiceNow spun up prompt engineering roles and required small evaluation datasets (e.g., 100 examples) before production to change QA practices.

ADVICE

Remove Token Caps Temporarily To Drive Adoption

Encourage broad internal experimentation by removing short-term token limits to break resistance and surface real usage patterns.
Jae ran a "Tokens Never Sleep" period and found per-person token use shockingly low, which accelerated adoption.

INSIGHT

Fix Data Structure Before Blaming Models

Poorly structured internal data, not the model, is the common cause of hallucinations and brittle agent behavior.
Jae and Krish stress fixing data pipelines and publishing discipline before blaming LLMs for errors.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

Executives from ServiceNow, TwelveLabs, and AWS share unfiltered lessons on organizational change, data strategy, and best practices for scaling AI products worldwide.

Topics Include:

Three panelists explore scaling and monetizing agentic AI products.
ServiceNow built entirely new prompt engineering roles from scratch.
Evaluations-first culture replaced traditional QA — a major mindset shift.
TwelveLabs' "Tokens Never Sleep" initiative broke down team AI resistance.
Unlimited token usage revealed how underused AI actually is.
Rich: the competitive window for acting on AI is closing.
ServiceNow's DART program governs customer data for evaluation only.
Enterprise data governance is fundamentally different from consumer companies.
Disorganized internal data breaks agentic systems — structure comes first.
Your data, not your model, is your competitive advantage.
Krish's surprise: enterprise governance made model training nearly impossible.
Jae: powerful AI tools erode human opinions faster than expected.
Rich: boardroom-to-shop-floor AI adoption is unlike any previous wave.
Customers began favoring velocity after watching competitors win with it.
ServiceNow shipped agents at 20% resolution — then iterated upward.
Focus early agents on reversible, low-risk, two-way-door actions.
AWS tracks internal AI adoption in a structured weekly mechanism.
TwelveLabs' two models power sophisticated video RAG workflows at scale.
A Hollywood studio cut full episodes to four minutes using them.
Voice agent worked perfectly in demos — broke immediately in production.
Customers now measure how fast a product is getting smarter.
Future-proofing infrastructure is every product leader's top anxiety today.
MCP and A2A enable message-passing — but deeper problems remain.
Semantic mismatch, agent identity, and trace governance remain unsolved.
Tiered autonomy, trust, and data foundations define who ultimately wins.

Participants:

Krish Ganapathy | VP, AI Science, Architecture and Tools, ServiceNow
Jae Lee | CEO & Co-Founder, TwelveLabs
Rich Geraffo | Vice President & Managing Director, North America, AWS
Moderator: Connie de Lange | Marketing Director, North America, AWS

See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/