Tomasz Tunguz’s recent article on selling AI and software struck a chord with me. As AI adoption grows, both buyers and sellers are facing new challenges—especially around scaling and discovery. Below are my thoughts on his key points, based on my live experiences selling the AI chatbot at Sendbird, and how they relate to what we’re seeing in the market. Check out the full article here: https://tomtunguz.com/software-playbook
Selling AI is being discovered because the technology is new. Buyers don’t know how to use it or how to buy it.
I fully agree with this and would emphasize it from two perspectives: selling to both enterprises and SMBs comes with discovery challenges. For enterprises, based on various studies and papers I’ve read, many buyers struggle with scaling their generative AI projects after initial adoption. They face issues expanding AI usage beyond a single application and unlocking its full potential while managing risks. For SMBs, there is still a significant technology and discovery gap. Similar to blockchain, the jargon and user interfaces of most generative AI applications remain too technical, and the product—as well as the introduction of the service—needs to be more user-friendly.
Because the sales motions are new, we can’t apply the previous playbook to the new sales process. The CEO/founder should hire a sales leader that they fully trust who focuses on ultimate success. The sales process is a part of the product.
I strongly agree that the sales process is integral to the product experience. Beyond self-serve onboarding, we must also offer an easy pathway for mid-market and enterprise users to engage with our sales or technical teams when needed. Even during the sales demo, we can start by showcasing a prototype chatbot tailored to the prospect’s website. This not only accelerates technical due diligence but also delivers a faster ‘wow’ moment. The sales and product experience should be tightly integrated.
Between PLG vs. sales-led, more companies were sales-led. If starting with PLG, the template sells the product. Fight the empty box problem with great concrete templates that demonstrate how to use AI. If selling top-down, most of the conversations today are at the C-suite rather than the mid-market predominantly because the buying process is new.
I firmly believe in this. Our platform, paired with strong templates, can effectively sell our AI chatbot product. For example, offering a user-friendly graphical interface for prompt templates instead of relying on text-based ones allows prospects to grasp the product’s potential without much technical effort. However, I remain cautious about top-down selling. If our SaaS product or PMF isn’t fully ready, customization efforts could disrupt our PLG motion. I’m not against taking on SI roles, but we must ensure we want to scale with our platform as a center of our growth
Figuring out how to consistently produce wow moments with non-deterministic software is essential.
Absolutely. This is a key focus for our product and marketing teams, and one of our OKRs this quarter is to improve conversion rates by driving consistent wow moments for our users.
The room was split on the pricing model: seats, usage, or some hybrid. Ultimately, pricing captures 15-30% of the value the software/AI creates. Developing a strong case for this with buyers will be key because the ROI question from buyers is real, especially as the broader software market feels pressure.
Value-based pricing is an excellent approach, but we also face perception challenges. Many customers still view AI products through the lens of past investments, like seat-based pricing models and ROI measured by traditional KPIs, such as cost per inbound/outbound call. We need to pioneer new value-based pricing anchors for AI, and while the challenges are significant, I’m ready to face them head-on.
Only VCs care about the word agents/agentic: for most enterprises, agents mean a customer support agent. Many teams don’t care about the underlying technology; they seek a solution to their problem.
Agreed. We’re still refining our ICP. We’re seeing enterprise leads from Heads of Data Science who have developed their own LLMs but now seek alternatives like Sendbird, where they can achieve faster value. At the same time, Heads of CS/CX are looking to solve business challenges but are frustrated by stalled internal projects.