We’re at the dawn of the agentic era. It’s early, but given the pace of AI, the entire enterprise will soon undergo its biggest shift since the introduction of PCs.

‍

‍

What will agent-first software companies look like in 5 years?

‍

Some predictions:

‍

🖥️ Product and Engineering  

• Everyone will be a developer, whether they are building customer-facing features or internal tools.
• Senior engineers will do little coding and instead focus on defining the rules of engagement: how to ensure code quality and consistency, how does agent orchestration work, how is data stored and secured, which do agents have access to, how are new features are tested and merged into production etc.
• The most talented engineers will continue to be in high demand, now freed up to work on the most challenging and impactful problems.
• ‍The lines between product and engineering will continue to blur. We'll see more and more forward-deployed engineers working on integration-heavy solutions. And we'll see PMs with clear understanding of customer needs ship features from idea to prod in hours themselves.
• All that will lead to a collapsing of the design, product and engineering silos allowing for more nimble and faster moving teams.
• ‍Software development costs will asymptote towards zero. Poorly solved edge cases today will have their own product solutions. Tailoring software to one or a few customers will be common.
• Agent product discovery will mine user behavior, market trends and competitor features to propose new features and products.
• Trial and error will be more efficient to grow revenue than product intuition. The primary bottleneck will be cohort sizes and customer attention for A/B testing...unless agents can accurately predict customer preferences.
• As code generation improves, new code architectures will emerge making agentic code maintenance easier and cheaper. Instead of senior engineers battling tech debt, the right code architecture will allow agents to continuously propose code improvements, further increasing development velocity.
• ‍Teams can 10x (or even 100x) product velocity if they optimize for critical human decision making - for example, strategic planning, data governance and approval processes.
• ‍Companies that have enough product whitespace to go after will continue to grow headcount in product and engineering. Companies that do not, will scale down headcount and optimize for efficiency.
• It will be common to see hyper-efficient startups generating $1m+ ARR/employee at 30%+ FCF margins (if they have a distribution moat).

‍

🤝 Sales and Marketing

• As product development costs plummet, distribution will become THE main thing.
• The power balance inside software companies will shift ever more towards sales and marketing, away from product and engineering.
• Margins will compress as the build cost in a build-vs-buy decision drops and software development get commoditized.
• Every company needs a LLM SEO strategy.
• We will see B2A(gent) businesses (in addition to B2B and B2C) that have no human interface.
• On the surface, B2A companies will look like developer tooling business models. But what if switching costs go away due to costless integration vs building against an API workflow? What will the moats be for B2A companies?
• Selling to agents will get optimized for / gamed like SEO today. (Will we see agents taking bribes to buy a product?)
• For complex, high ACV enterprise sales, human-to-human selling will not get replaced any time soon. Although initial contact points will be automated.
• For lower ACV SMB sales, agent-human, human-agent and agent-agent sales will become common.
• ARR and new ARR per sales rep will increase significantly as a result.
• But increases in sales efficiency will get offset by an increase in marketing costs as marketing channels get flooded by new software offerings.

‍

💬 Customer success

• CS agents will proactively predict churn risk and automatically implement retention strategies
• We will see near instantaneous, automated feature deployment and bug fixing: going from customer complaint → bug documentation → code fix → test → merge into prod…all with minimal human oversight.
• Human involvement in customer support will be limited to complex problems and relationship management

‍

➕ Bonus predictions

• Agents will continue to improve at whiplashing speeds. Teams that adapt faster will have a huge competitive advantage in the transition phase towards the Agentic Enterprise. Adaptation speed as a wedge to win market share.
• We will see the equivalent of SaaS vending machine businesses: once set up, they operate fully autonomously and pay its owners a passive income.

‍