Co-Founder Skills, Building Gong, And What PMs Do Best

Harshal
1 day ago
7 min read

Insights From Eilon Reshef (Co-Founder And CPO, Gong)

I spoke with Eilon Reshef, co-founder and CPO at Gong, some time ago, and kept coming back to some of the learnings. I'm using this post to share my learnings from that conversation. Eilon shared about co-founders, product, AI, and Product Management. Gong is a deep-tech, AI-powered B2B product built for sales. In this conversation, he shared practical lessons on how to pick co-founders, build deep-tech products, use untapped data, choose a GTM motion, and more. This article distills those lessons into clear, actionable ideas.

I spent 1 hour 35 minutes writing this. You need 7 minutes to read this.

Eilon with illustrations of going from data to graphs and leveling up.

Why This Conversation Matters

Founders face similar problems: choosing the right mix of skills in the founding team, deciding how deep the technology must be, and understanding real customer needs instead of building from guesses (a challenge for PMs too). Deep-tech and AI products are especially hard because they rely on new or underused data and must turn noisy raw data into useful, explainable insights. Eilon shows how to start narrow, learn from real usage, and align sales motion, product choices, and data strategy. The rest of this article breaks those ideas down into specific decisions you can reuse as a Product Manager or a Founder.

1. Finding Startup Co-Founders

Pick two or three co-founders with complementary skills across product, engineering, sales, and marketing so you can move fast and make clear decisions. At the start you need sales more than marketing; after roughly five B2B customers, marketing starts to matter. It is common to see two co-founders - one covering product and engineering, one covering marketing and sales. Do not start with 4 co-founders. Two or three is better, to improve decision-making velocity.

Founding Sales also highlights the importance of founders selling the product.

For example, Eilon and Amit (CEO) split responsibilities this way. Eilon was the de facto CTO at first and coded the first version of Gong. Later he hired a head of R&D and moved fully into product while Amit owned sales and marketing. That split - product/tech on one side, sales/marketing on the other - worked and stayed stable as the company grew.

Meeting a co-founder is a bit like meeting a spouse: luck and timing matter and there is no fixed playbook. Eilon met Amit through a mutual friend during a sabbatical after selling a startup. It took more than one meeting to decide to work together, which is a reminder that co-founder decisions are built over repeated conversations, not one moment.

2. Deep Tech Products

Be honest about how deep the technology in your product really is. If you are reasonably technical, you can often build a B2B SaaS MVP yourself. You can put up a landing page, stitch together tools, and ship something basic to real users.

But if the product is truly deep-tech—like Gong’s use of NLP on sales conversations - you need a strong technical founder to own that area. Deep-tech products depend on real advances in data collection, modeling, and error-rate reduction, not just glue code around existing tools. In that case, treat “deep technical leadership” as a founding skill, not an afterthought or a paid hire.

It is also better to avoid hiring a full development team before you have funding and validation. Early on, keep the team small, use your own skills to learn quickly, and save the heavy hiring for after you know customers care enough to pay.

3. Take Sabbaticals

Use sabbaticals as an intentional reset between big chapters. Eilon recommends taking a sabbatical after selling a company or ending a major chapter because new chapters often start there. Time away from execution creates space to think, explore, and meet people without the pressure of a roadmap.

During his sabbatical, Eilon took many meetings, and one of those meetings led him to Amit. Gong came from that period. The pattern is simple: close a chapter, take time off, meet many people, and let new ideas and relationships emerge instead of jumping straight into the next job.

4. Luck And Data For Product-Market Fit

Aim for product–market fit by combining a real, felt pain with luck and a new data source.

For example, Amit, Gong’s CEO, had already seen the pain: despite having a CRM, leaders lacked visibility into why sales were low. That context helped them recognize a problem that had a willing target market. Gong also uncovered a new data source: sales calls were not being recorded or analyzed before. Many AI startups only analyze existing data like email, where everyone has access and the moat is thin. New or underused data (such as phone call audio at Gong) is where much of the opportunity lies, because it lets you see behavior others cannot.

5. Vertical Focus

Resist the temptation to build for many verticals at once. Choose one problem, one vertical, and a simple first product, then use adoption and willingness to pay as your guide.

For example, the first version of Gong was simple: record, transcribe, comment, and share sales calls. It focused on integration with WebEx, because this was before Zoom, and targeted sales teams. They gave this early product away for free to businesses to learn from real usage and to collect data. This let them watch how people used the product, improve the experience, and validate that the problem was real. When they started charging, 11 of 12 (formerly-free) customers paid, which was a strong signal that the solution to the narrow problem in the narrow vertical were working.

6. B2B Top-down Sales

Do not be afraid to go top-down in B2B if the buying process demands it. Product-led growth and bottoms-up adoption make for great stories, but many companies still make a lot of money every year through traditional enterprise sales. Even very product-led companies like Twilio or n8n still hire sales teams. The pattern is: let the product be strong and easy to try where possible, but design your motion around how buyers in your space really purchase, not just around idealized PLG stories.

For example, Gong chose a top-down motion with sales as the entry point into organizations. Legal and compliance often care whether calls are recorded, so a single salesperson cannot always just sign up and use the product. You may need to win over leaders, legal, and security before usage can spread.

7. Data To Wisdom

Treat data as a ladder from raw signals to practical wisdom, and be clear about where your product lives on that ladder. Frame it as Data → Information → Knowledge → Wisdom. Not all data is informative, and not all information is valuable on its own.

For example:

A phone call transcript is data.
Topic detection on that transcript is information.
Recommendations about what to do based on those topics are wisdom.

Predictions sound exciting, but business users often care more about understanding than about abstract predictions. They want to know what happened, why it happened, and what to do next in a way they can trust and act on.

For example, One of Gong's AI products had limited success because the recommendations were hard to explain and hard to act on in real time. Gong instead focused on reducing time to consume information (for example, by ~50x with topic detection) and steadily improving NLP error rates from around 30% to around 10%. They trained on sales conversations and CRM context, such as customer name, not on generic academic datasets like reviews. Gong also chose not to overload reps with too much real-time feedback because it can distract them in the moment. Across all of this, the goal was efficiency and better decisions, not replacing people. The product succeeds when it turns noisy, hard-to-use data into clear, trustworthy insight that fits into how sales teams already work.

8. PMs = Customer-needs

Make “understanding customer needs firsthand” the core of the PM role. For example, every new PM should talk to 10 customers before starting any work. The most valuable part of product management is customer needs discovery, alongside the broader responsibility of helping build the right product.

PMs should not rely only on summaries. If a customer says something important in a call, AI may tag it. But instead of just reading the tags, PMs can go directly to the source recording rather than hearing a second-hand explanation. This creates a steady stream of information from real usage into the product team.

I captured some more tips on Navigating The First Product Manager Role At Startups.

The pattern is: structure your org so PMs keep talking to customers, listening to raw calls, and grounding prioritization in lived customer pain rather than internal debates or slides.

9. Target Budget-holding ICPs

Choose an Ideal Customer Profile that both feels the pain and holds a budget. When building a startup, it is not enough that a group of users benefits from your product; they also need to be able to buy it.

For example, sales and marketing leaders often have a budget for new tools, while PMs usually do not hold the same kind of buying authority. An enterprise SaaS startup often needs to make around $1,000 per seat annually. Companies can justify that for sales productivity, but they rarely can for PM productivity in the same way.

For example, Gong does not sell to PMs, even though the product helps PMs too. The paying customer and ICP is the sales organization. The lesson is to align your product, messaging, and go-to-market with budget-holding personas, not just with the people who benefit indirectly.

10. Write In public

Write in public so people can see how you think before they work with you. Writing and public thinking help when people evaluate you, especially in high-stakes transitions. If your startup is being acquired, buyers will research the people involved and look for public thinking. Having your views and writing out there helps during due diligence because it makes your reasoning style and values visible. Public writing is also a forcing function. It pushes you to clarify your own thinking on products, markets, and decisions, which in turn makes you a better founder, PM, or product leader.

What Stayed With Me

From this conversation with Eilon, a few themes stood out to me. Co-founding teams work best when they are small, complementary, and clear about who owns product/engineering versus sales/marketing. Sabbaticals and prior experience create the space and context that lead to the next chapter, including spotting real pain and being ready when luck arrives.

The data-to-wisdom ladder is a useful lens: focus less on flashy predictions and more on turning new, underused data into explanations that save time and improve decisions. Along the way, PMs should own direct customer understanding, ICPs should be budget holders, and consistent public writing should show how you think.