Deepti Dheer

Situation

Developers wanted to build and run AI agents over their own data but were stitching together brittle tooling. Elastic's goal was to make its platform agent-centric and capture that demand with a first-party, monetizable agentic experience that stayed cost-efficient at scale. I owned the vision, strategy, and roadmap from a blank page.

Task

• Define product vision, strategy, and roadmap from scratch
• Get a credible product to GA fast, then to first revenue
• Build pricing and packaging for hosted and pay-as-you-go billing
• Keep agent token costs sustainable as usage scaled
• Drive adoption through a product-led-growth motion
• Align 7+ cross-functional teams on one unified agentic experience

Action

I scoped aggressively to reach GA in three months, then prioritized monetization over breadth, shipping hosted and pay-as-you-go billing before chasing every feature. I partnered with engineering and data science to architect a context engine with dynamically loaded skills, a conversation context store, and selective compaction, cutting token cost rather than simply adding capability. I set a product-led-growth motion that embedded Agent Builder into customers' own tools through Slack agents, Claude Code, and Cursor over a 30+ connector network, and unified more than seven teams' efforts into a single agentic experience across Elasticsearch, Observability, and Security instead of fragmented per-product agents.

Situation

More than 100 million customers needed trustworthy, in-product help across complex financial and tax tasks, but generic LLM answers were not safe or accurate enough. The business goal was to contain support demand, raise satisfaction, and grow revenue with a GenAI assistant that worked across web, mobile, text, and voice without hallucinating in a high-stakes domain.

Task

• Pioneer the zero-to-one GenAI growth and support strategy
• Lead a 40+ member cross-functional team from ideation to launch
• Make responses accurate and safe enough to trust in a regulated domain
• Contain support demand and lift customer satisfaction
• Scale the assistant across surfaces and business units

Action

I treated response quality as the product rather than coverage. I partnered with data science to enrich models with tax and accounting knowledge through RAG and Graph-RAG retrieval, ranking, and attribution, and refined NLP intents and tone sensitivity with AI safety features such as proactive profanity filtering and bias detection, accepting a slower rollout in exchange for trust. I automated in-product routing and ticket creation to lift service efficiency, drove the experience through rigorous A/B/C experimentation rather than intuition, and scaled across surfaces and three business units only after the core experience had proven out. The CEO unveiled the launch at the Investor Summit.

Situation

Retained B2C and B2B customers were under-monetized and receiving generic experiences, and product teams lacked a rigorous way to test what actually moved revenue. The goal was to lift recurring revenue through personalization and disciplined experimentation, surfacing the right next product or action without degrading the experience with irrelevant prompts.

Task

• Build a zero-to-one ML recommendation engine for cross-sell and content
• Improve personalized targeting and customer segmentation
• Stand up rigorous experimentation to find what moves revenue
• Protect experience quality and model trustworthiness while scaling

Action

I prioritized ranking and retrieval quality and built a Data Enrichment platform capability to improve targeting and segmentation, gating cross-sell on relevance signals rather than maximizing surface area. I partnered across design, research, legal, compliance, marketing, engineering, and data science to develop and test hyper-personalized experiences on web, mobile, text, and voice, and I designed A/B tests with statistical rigor to translate complex findings into product improvements. I also simplified the in-product search experience with AI, and built early AI testing frameworks to evaluate model performance and surface bias before scaling.

Situation

QuickBooks users could schedule reminder emails to chase unpaid invoices, but the process was manual and time-consuming, and many invoices were still paid late or not in full, straining small-business cash flow. The goal was to make getting paid effortless by letting AI handle the chasing end to end.

Task

• Remove the manual effort of chasing overdue invoices
• Use an AI agent to detect, draft, send, and follow up automatically
• Keep the business owner in control with human-in-the-loop approval
• Meet UX, architecture, legal, and compliance requirements
• Build a reusable AI workflow other product SKUs could adopt

Action

I designed an AI agent that continuously monitors invoice status and decides when to act. GenAI drafts a personalized, contextual reminder from the customer's payment history and surfaces it to the business owner, who can approve or regenerate before it sends, keeping a human in the loop. If the payer does not respond, the agent follows up with a GenAI escalation message that adapts its tone and content, and once payment lands it notifies the owner. I partnered with engineering, design, and compliance to finalize the UX, architecture, and legal requirements, and stood up a delivery cadence to ship it. The work was featured on Intuit's CTO blog.

Situation

Enterprise prospects had no low-friction way to evaluate products, and support could not scale to the growing volume of interactions, all inside a traditionally sales-led organization. The goal was to prove that a self-serve, AI-assisted motion could generate real pipeline, and to quantify the revenue gap in order to win executive buy-in.

Task

• Quantify the revenue gap and secure C-suite funding
• Build a zero-to-one freemium trial marketplace for NetApp and partner products
• Feed self-serve demand into the existing sales pipeline
• Ship an AI chatbot reliable enough for operational scale
• Expand chatbot capability and lift subscription renewals

Action

I quantified revenue gaps worth $380M and took business plans to the C-suite to fund the bet. I built the freemium trial marketplace from scratch, spanning NetApp and partner products from AWS, GCP, Azure, HP, Hitachi, and Dell-EMC, and integrated it with Salesforce CRM and marketing automation so that self-serve fed the existing sales motion rather than competing with it. I shipped a zero-to-one AI chatbot and optimized its NLP pipelines for operational scale, reaching 10,000 daily interactions at 95% intent accuracy, choosing reliability over breadth. I then iterated rapidly with design and engineering to add order tracking, reminders, notifications, and automated re-order placement, and led a global team on a go-to-market strategy that lifted subscription renewals.