How Deel's product marketing team turned tens of thousands of customer calls into a competitive edge
Deel closed its Customer Context Gap — turning Gong calls, CRM data, reviews, and support tickets into an intelligence layer that tells product marketers what's winning deals, what's losing them, what product bundles to create, and what the messaging should be.
The challenge
Deel is one of the fastest-growing companies in HR and payroll, scaling rapidly through organic growth and acquisitions in a fiercely competitive market. In a space where customer focus makes the difference, understanding what customers say, need, and want isn't a nice-to-have — it's a strategic edge.
AI is compressing work cycles like never before. Teams have more choices, need to make more decisions, and must execute in shorter timeframes. In this context, staying close to customers is critical. And Deel has more customer signal than most: tens of thousands of Gong calls per month across sales and customer success, plus reviews on G2, Trustpilot, and the App Store, notes in Salesforce, tickets in Zendesk, and more.
But having the data isn't the same as using it. Deel's dashboards could show what was happening — deal velocity, churn rates, product adoption metrics. What they couldn't explain was why. Why are enterprise prospects choosing Deel over a specific competitor? Why is messaging resonating in one region but falling flat in another? Why do certain products get mentioned together in sales conversations?
The answers lived in what customers and prospects were actually saying in calls and reviews — but at Deel's volume, no human team could read, tag, and synthesize it fast enough. Product marketers needed access to this intelligence to drive growth, sharpen positioning, and keep Deel ahead in the market. Instead, they were operating on a fraction of the available signal — and the rest went unused.
Results at a glance
From scattered signals to one intelligence layer: Customer calls, reviews, CRM data, and support tickets are continuously processed into an interpreted, structured layer — not raw dumps, but reasoned outputs product marketers can act on immediately.
Token-efficient, instant answers: Because NEXT AI builds an interpreted intelligence layer from the raw data, every query is fast, reliable, and cost-effective — no context-window overload, no running out of tokens mid-analysis. A frustration teams across industries know all too well with general-purpose AI tools.
Evidence-backed GTM execution: Messaging, positioning, battle cards, and launch decisions are grounded in quantified customer evidence — not internal assumptions or a handful of cherry-picked quotes.
Broader adoption in motion: Starting with a core group of PMMs, now expanding across all product marketing, with a path to marketing leadership and growth teams.
Inside Deel's Customer OS
Deel uses NEXT AI as its Customer OS — the intelligence layer that bridges the gap between transactional data and real customer context:
Unify — Gong calls (sales and success), G2 and Trustpilot reviews, App Store reviews, Salesforce notes, and Zendesk tickets flow into one system, organized around the products, segments, personas, regions, and deal outcomes behind them.
Understand — AI agents continuously process incoming conversations and feedback: clustering themes, quantifying what's driving wins, losses, churn, and adoption — surfacing patterns, objections, and opportunities across the full dataset, not just keywords.
Act — Intelligence connects back into the flow of work: messaging frameworks validated against real language, battle cards that update themselves, persona-specific positioning grounded in evidence, and structured reports that used to take a week to build.
Connected sources:
Gong (tens of thousands of sales and success calls per month)
Salesforce (CRM context: segment, industry, ARR, region, deal outcome)
G2, Trustpilot, and App Store reviews
Zendesk (support tickets)
Deel's signature NEXT AI workflows
1) Messaging and competitive intelligence — know what wins and why
The job: Validate positioning, sharpen messaging by persona and segment, and keep competitive battle cards grounded in what's actually happening in the field — not desk research that's outdated the week after it's published.
How NEXT AI does it: PMMs validate messaging frameworks against real customer and prospect language from recorded calls — removing guesswork and grounding decisions in what customers actually say. NEXT AI surfaces which features are most frequently linked to won deals ("Why We Win"), maps the exact objections prospects raise by product, segment, and deal stage, and tracks which competitors come up most often and what specific claims prospects repeat. AI agents continuously process incoming calls, so battle cards and competitive views stay current without anyone manually listening to recordings.
The result: Messaging that mirrors how customers actually talk. Battle cards that reflect the real competitive landscape, not a quarterly snapshot. Positioning validated against evidence, not assumptions.
2) Product intelligence — ground roadmap influence in customer evidence
The job: Give product marketers a quantified, always-current view of how customers experience each product — what's resonating, what's causing friction, and what's being requested — so they can influence roadmap and launch decisions with evidence, not anecdotes.
How NEXT AI does it: PMMs ask questions like "What are the top customer challenges for Deel Payroll in the past 90 days?" or "Which feature requests are rising fastest by segment?" NEXT AI reasons across all the call and review data — not a sample — and returns ranked drivers with counts, representative quotes, and segment breakdowns. After a launch, PMMs can track reception in real time: what's resonating, what's confusing, what's being ignored. NEXT AI also reveals which products tend to be mentioned together in conversations, surfacing natural bundling patterns and cross-sell opportunities.
The result: Structured, evidence-backed reports that replace hours of manual call review. Roadmap influence grounded in quantified themes, not the loudest opinion in the room. Launch feedback loops that don't wait for survey data.
3) Customer pulse and decision validation — pressure-test everything
The job: Maintain a continuous read on customer sentiment across the portfolio, spot shifts early, and pressure-test GTM bets before committing resources.
How NEXT AI does it: NEXT AI can auto-generate quarterly executive summaries combining sentiment trends, emerging themes, competitive shifts, and product perception across all call data — replacing the manual "state of the customer" deck that takes a week to build. PMMs also validate hypotheses directly: "Is this persona responding to our new positioning?" NEXT AI surfaces supporting signals, counter-signals, and segment differences, grounded in evidence across the full dataset. For churn, NEXT AI analyzes the full call history of churned accounts to identify early warning language and the specific unmet needs that drove departures — giving teams a quantified view of why customers leave, not just that they leave.
The result: A living customer pulse that leadership can trust. Fewer mis-bets on campaigns, messaging, and positioning. Churn intelligence that's actionable — including prioritized win-back candidates where Deel has since addressed the original pain point.
Why NEXT AI — and why not just build it or use ChatGPT/Claude
Deel's teams already have access to leading AI models — and had already invested in building internal AI agents. The engineering capability was there. But two things became clear.
General-purpose AI hits structural limits at scale. Pasting call transcripts into a chat window means working with raw, unstructured data in a limited context window. At Deel's volume — tens of thousands of calls per month, plus reviews, tickets, and CRM data — that approach breaks down fast: token limits cut you off before you see the full picture, answers vary from prompt to prompt, and a few quotes can sound like a universal truth. Teams across the industry know the frustration of running out of tokens mid-analysis or getting a different answer every time they ask the same question.
Building the last 20% is where the real cost lives. Getting an AI prototype to 80% is fast. The last 20% — retrieval accuracy, governance, token efficiency, evaluation frameworks, model churn management — quietly consumes 80% of the effort and budget. And the cost doesn't stop at v1: regression testing on every model upgrade, tuning retrieval as data scales, managing token costs that spike unpredictably, and maintaining connectors as source systems evolve. Every month spent on internal plumbing is a month not spent on shipping product differentiation. This mirrors a broader market shift: according to Menlo Ventures' State of AI report, 76% of enterprise AI use cases are now purchased rather than built internally, up from 47% just two years ago.
Deel chose to deploy NEXT AI as a managed Customer OS — getting to production-grade customer intelligence fast, while keeping engineering focused on what makes Deel's own product win. Here's what that means in practice:
Interpreted intelligence layer: NEXT AI continuously ingests, cleans, and structures raw conversations and feedback into a governed data model — so every query draws from reasoned, organized customer intelligence, not raw text dumps. This is what makes it token-efficient, fast, and reliable at enterprise scale.
Reasons across all the data: Not a sample, not a list of verbatims. NEXT AI quantifies themes, ranks drivers, and surfaces evidence across the full dataset — producing outputs teams can actually act on: ranked priorities, validated messaging, structured reports.
Model-agnostic and future-proof: NEXT AI works across AI providers (OpenAI, Anthropic, Google) and absorbs model churn — regression testing, prompt tuning, and cost controls — so teams don't have to rebuild every time a model updates.
From insight to execution: Battle cards, messaging frameworks, persona analyses, launch trackers, and executive summaries connect directly back into the flow of work — for humans and AI agents alike.
Enterprise governance: Evidence-first answers with full traceability, permissions, audit trails, and PII protection built in.
Leadership lessons from Deel
Customer focus is a competitive edge. In a crowded market, the companies that understand their customers most deeply — and act on that understanding fastest — win. The Customer OS is how you operationalize that advantage.
Actionable beats comprehensive. A list of 500 call excerpts isn't an insight. Ranked drivers with evidence, segment breakdowns, and clear direction — that's what moves a PMM team from reactive to strategic.
Buy the infrastructure, build the differentiation. The 80/20 trap is real: getting AI to "kind of work" is fast; making it reliable, governed, and cost-efficient at scale is where the real investment lives. Smart teams buy that layer and focus engineering on what makes their own product win.
Speed compounds. When PMMs can validate messaging, build battle cards, and generate reports in minutes instead of weeks, the entire GTM engine accelerates.
Looking ahead
Deel is already using NEXT AI as its Customer OS for product marketing — the intelligence layer that bridges the gap between the volume of customer conversations Deel captures every day and the GTM decisions that need to be grounded in what customers actually say. As adoption expands from PMMs to marketing leadership and growth teams, Deel is building toward a single, shared customer truth that powers every GTM motion — messaging, campaigns, competitive plays, churn saves, and roadmap influence — across products, segments, and regions.