Reasonable Predictions, Lightly Held
Some working hypotheses for this year, and what might follow.
Early February feels like a reasonable time to make imperfect predictions. New Year optimism has worn off, the resolutions are fading, and the systems are already pushing back.
Every year around this time, we compare notes inside Intentional, and with friends, clients, and peers. We ask: what’s already happening, just not evenly yet? What’s being overhyped? What’s quietly not working? What’s about to become obvious? Well, we got answers!
This is a meaty post so take your time, and let us know what you think in the comments. We love to hear dissenting opinions.
Leadership & Organizational Clarity
Your hero’s journey will matter less than having an answer to “can you just explain what’s going on?”.
Sihan Shi, Product Consultant @ Intentional: As companies rush to adopt AI, the existing gaps in their company strategy, team structure and culture will be exacerbated. But rushed adoption is going to mask those original sins and make it harder for leaders to diagnose and address them.
Neha Thanki, Product Consultant @ Intentional: Clarity will outperform charisma in leadership. In 2026, leaders who are decisive, explicit, and sometimes boring will outperform those who inspire without resolving tension. Perhaps this is common sense (or maybe manifestation?) but I see AI accelerating this shift. Working with systems that demand explicit intent is training teams to expect the same clarity from people.
Go-to-Market
We’ve automated distribution, but not attention 🙃
Dexter Shepherd, Product Consultant @ Intentional: In 2025, GenAI broke cold outbound. The marginal cost to send customized outreach dropped to near zero, inboxes were flooded, and prospects completely checked out.
The same is going to happen to B2B content in 2026. Feeds and inboxes of your target customers will be flooded with (even more) content, and it’s going to be increasingly difficult for them to discern the slop from the sublime with the limited time and attention they have.
Unless companies find genuinely compelling ways to provide value, readers will get overwhelmed and disengage. Great time to start a newsletter, amiright!?
Jason Yang, Co-founder @ Andoe: In 2026, companies will realize that AI answers are now their first impression, and most of them won’t like what they see. As AI-generated answers become a primary interface for information, brands will be introduced, summarized, and compared long before a website visit or sales conversation ever happens.
Over the past year, I’ve seen teams chase AI visibility “hacks”: keyword-stuffed pages, synthetic FAQs, prompt baiting, and content written for models instead of people. In 2026, those shortcuts will backfire not as ranking drops, but as omissions.
The shift will be from hacking toward treating AI visibility as a core strategic signal to guide positioning, content, and brand decisions. Teams will focus on showing up correctly across two major AI providers, and let everything else follow.
Robbie Larson, Product Lead @ Coconut: GPT/perplexity apps will launch in 2026 but they will bomb because they haven’t and won’t figure out how those apps are discovered in a chat interface
Sharva Hassamal, Founder @ Portage Labs: In 2026, the GTM middle gets hollowed out. Scrappy teams move faster by stitching together low-cost, AI-native tools and agentic workflows, while enterprise doubles down on fewer, deeper platforms that actually execute decisions, not just generate ideas.
For small and mid-size teams, this makes traditional SaaS stacks feel expensive, rigid, and overbuilt—why pay for a bloated CRM when you can spin up something custom in weeks? At the same time, incumbents embedded in enterprise gain an edge by layering AI on massive datasets, with platforms like Adobe, Braze, and Databricks turning agentic workflows into a core feature rather than a bolt-on.
Stephen Trim, GTM Sales Engineer & Partnerships @ SnapPea: In 2026, the BDR role will start to materially fade away. Early stage companies will move directly from founder selling to Agents, AEs, Marketers, and CS for Rev OPs workflows
Zach Vidibor, Co-founder & CEO @ Octave: Despite all the doomsayers that said AI was going to kill sales roles, I’m positive there will be more sellers, and sales job openings, this year than last. Someone has to sell all this code that’s so easy to create now, right 🙂
Product, Discovery & Judgement
We can ship faster than ever but figuring out what actually matters remains a very human problem.
Nathan Ngai, Founder & CEO @ Arkhet: When building becomes cheap, knowing what to build becomes priceless. But we’re about to learn this lesson the hard way.
AI is making code so accessible that PMs and designers are being pushed to ship directly to production. Not as a superpower for discovery, but as a replacement for engineering headcount. Output is measured in commits instead of customer impact.
The result will be a wave of well-built products that nobody uses: pristine code, slick interfaces, and zero adoption. Not because teams lack talent, but because discovery is being skipped entirely. Getting better at AI doesn’t make you better at discovery. These are different muscles. Teams who use AI to learn faster will outperform teams who use AI to ship faster.
Steven Chen, Staff Product Manager @ KitBash3D: Claude Code adoption amongst non-technical users’ barriers will be lowered further. General purpose agents will become more common outside of engineering and more GUIs will start to be built on top to make a terminal interface less scary.
Deciding what the work is and reviewing what work is done will become even more critical. The bottleneck moves from “building the thing” to “are we building the right thing?” and “is this built well enough to solve the problem?”
Existing mental models and UX patterns will need to be rethought to account for agentic usage and people interacting with software in fundamentally different ways. APIs and technical design will increasingly overtake UI patterns as the gold standard for defining product quality.
Agents & Infrastructure
Software is starting to behave differently. Less obedient toddler, more moody teenager perhaps.
Dan Schissler, Founder @ ConSalt: Frontier AI labs will achieve meaningful performance gains by introducing architectures inspired by human sleep and memory consolidation. Separating learning, abstraction, and long-term memory into distinct phases will prove to drive model improvement rather than relying on continuous training alone.
Quin Sandler, CEO @ Plantiga Technologies Inc.: Everyone is talking about managing agents. In 2026, the real shift won’t just be people using agents—it’ll be agents talking to other agents to get work done.
We’ll see a fundamental change in how systems connect. APIs won’t disappear, but the conversation will expand to MCPs (Model Context Protocol) as companies race to make their data and services agent-accessible. If your system can’t talk to an agent, you’re leaving opportunities on the table.
Proof of Work
Work is changing faster than the systems built to evaluate it.
Genveviève Smith, Partner @ Fieldwork: In 2026, candidates and hiring teams will finally acknowledge that existing hiring systems are fully broken. We may start to see the first real attempts to fix them, too.
We’ve been thinking about this through the lens of procedural debt. AI is enabling so much process that we are reaching a point where bots are talking to bots, agents are talking to agents, and no one is quite sure who the system is actually for anymore.
Stephen Southin, Founder @ Pave: Before EoY, people will start listing their preferred agentic stack alongside skills on their resumes. Agent choice and orchestration will become a visible marker for how people actually get work done.
Risk & Security
We’ve made everything easier, including the wrong things.
Tom Walsham, Principal & Founder @ Intentional: As cybersecurity leaders struggle with the ambiguous threat of shadow AI, 2026 will see the rise of a more tangible attack vector: the release of an open source LLM purpose-built for vibe hacking, with malicious alignment, and agentic approaches to probing targets and exploit chaining. Like the script-kiddies and self-propagating worms of the early 2000s (the bane of my life when I was a sysadmin early in my career), cursor-style IDEs will help less-technical hackers to exploit websites, unsecured IOT devices, etc., for fun & profit!
Genveviève Smith, Partner @ Fieldwork: By the end of 2026, deepfakes will be so easy to produce that we’ll need personal Turing tests for job interviews. “Am I actually talking to a real person?” will become a normal question. We may even need safe words for family phone calls, because this cannot be my brother asking me to wire bail if he didn’t say pineapple.
Thank you to the many smart, generous people who shared their thinking and let us include it here.
We’ll revisit this list in December and see what held, what aged badly, and what became obvious in hindsight. Until then, the conversation is more interesting than the scorecard so share what you’re seeing or tell us what we missed in the comments below!