AI-generated pitch decks hit a wall with venture capital in 2026
The AI-generated pitch deck has become the default output of modern startup fundraising, and it is starting to backfire. In a LinkedIn post published six days before 18 August 2026, communication coach Jonathan Millard claims that “90% of pitch decks and investor outreach are now AI-generated”, and that VCs are tuning out as a result. The issue, he argues, is not that the technology is poor. It is that the language is too polished, too interchangeable, and therefore emotionally flat. In other words, it reads like it came from a machine, because it did.
That single observation captures a broader shift playing out across the venture ecosystem. Founders have more tools than ever to generate decks, emails, scripts, and even Q and A prep. And yet, the conversion rate from outreach to meeting, and from meeting to conviction, is not simply a function of “better copy”. Investors are increasingly filtering for something harder to automate: lived experience, clarity under pressure, and the ability to make a room feel the opportunity (not just understand it).
And this is where the story gets interesting. The AI-generated pitch deck is not going away. If anything, accelerators and founder programmes are actively normalising AI in the company-building workflow. But the bar is moving. The winners are not the founders who use AI the most, they are the ones who use it without sounding like everyone else.
What the headlines are really saying: VCs are bored of “perfect” decks
Millard’s post is blunt about what investors are hearing on repeat: “Massive market.” “Disruptive platform.” “Strong team.” These phrases are not wrong. They are just empty when they arrive without specificity, without tension, and without a human being who clearly believes every word. Millard’s central claim is that correctness does not create conviction, and that the pitch is “not won on the page alone”. It is won when the founder speaks with belief, clarity, and verbal command.
That distinction matters because it reframes the fundraising problem. Many founders treat the deck as the product, and the meeting as a formality. But in practice, the deck is often a ticket to a conversation, and the conversation is where risk is assessed. Investors are not only evaluating the market and the model. They are evaluating the operator. Can this person recruit? Can they sell? Can they handle bad news? Can they learn fast? A deck can hint at those things, but it cannot prove them.
There is also a structural reason the “AI polish” is failing. When outreach becomes cheap to produce, volume goes up. When volume goes up, attention becomes scarcer. And when attention becomes scarcer, anything that looks templated gets deprioritised. It is not personal, it is triage. The irony is that founders adopt AI to stand out, then end up blending into a sea of near-identical narratives.
How the AI-generated pitch deck became the default fundraising artefact
The modern pitch deck has always been a template-driven format. Tools and libraries encourage best practice: problem, solution, market, traction, team, ask. Figma’s pitch deck resource library, for example, lays out what a strong deck should include, from mission and vision through to traction and financial projections, a funding request, and a call to action. It also offers a catalogue of examples and templates designed to help founders “tell the story” as much as present the idea. That is sensible advice, and it reflects how investors actually consume information.
But templates have a downside. They standardise the narrative arc. Then AI arrives and standardises the language inside that arc. The result is a deck that is structurally correct and linguistically smooth, but strategically weak. It does not have sharp edges. It does not have a point of view. It does not sound like a founder who has spent months talking to customers and getting punched in the face by reality (which, frankly, is what investors want to hear about).
Founder education is also accelerating this shift. The Founder Institute’s flagship FI Core programme explicitly includes setting up an AI toolkit, proving AI proficiency, and following an Agentic Track, described as an AI-personalised curriculum that adapts to the founder’s stage and includes “powerful free AI agents” to help complete work. Early deliverables include drafting first pitches, building mockups, testing pricing, and preparing for milestone reviews where founders are rated on idea viability, presentation skill, and research thoroughness. In other words, AI is being baked into the process, not bolted on.
So the AI-generated pitch deck is not an accident. It is the predictable outcome of a startup world that prizes speed, iteration, and repeatable playbooks. Fair enough. But when everyone uses the same playbook, differentiation moves elsewhere.
Jonathan Millard and the return of “oracy” as a founder advantage
Millard positions himself as someone who has “worked across five arenas: from boardrooms and radio studios to political forums and live events”, and he argues the pattern is consistent: people remember the person who sounds certain about what they are building, not the person with the most impressive algorithmic wording. His call to action is a “Voice Audit” and a free copy of “Oracy Mastery”, hosted at oracymasters.co.uk. The commercial angle is obvious. But the underlying point is still worth taking seriously.
Oracy, the ability to speak well, is not just presentation polish. In fundraising, it is a proxy for thinking. A founder who can explain a complex product simply, handle interruptions, and answer uncomfortable questions without spiralling is signalling competence. And competence is what investors buy when the numbers are still early and the product is still evolving.
Millard also makes a nuanced claim that often gets lost in AI debates: AI can help structure thinking and sharpen copy, but it cannot replace lived experience, risk, conviction, or voice. That is a useful framing because it avoids the lazy “AI bad” conclusion. The problem is not AI. The problem is founders outsourcing the parts of the pitch that should be most personal, most specific, and most earned.
Inside the pitch deck template economy: what Figma’s examples reveal
Figma’s resource library is a window into how pitch storytelling has been codified. It describes the pitch deck as “often the first real look someone gets at your idea”, and stresses that “how you tell the story matters as much as the idea itself”. It then lists the core components of a strong deck, including market overview, traction, team, and funding request. None of this is controversial. It is the baseline.
Where it becomes revealing is in the examples. The library points to Airbnb as a “classic example” of starting strong by addressing key pain points, then reinforcing the pitch with press mentions and positive user experiences. It also notes a specific target: Airbnb’s goal of 80,000 transactions within a year. That detail matters because it is concrete. It is measurable. It gives the investor something to react to, and it gives the founder something to be held accountable for.
Another example highlights Figma’s own early fundraising story: before having a working product, founders Dylan Field and Evan Wallace used a pitch deck to raise $3.8 million in pre-seed funding. The library says they emphasised what made Figma unique and focused on three core principles for online creative tools: access, building community, and education. Again, the lesson is not “use nicer words”. It is “have a clear thesis”. AI can help articulate a thesis, but it cannot invent one that stands up to scrutiny.
And there is a subtle warning here for 2026 founders. Templates are helpful, but they can also push founders towards generic claims. If a deck is built from a template and then filled with AI-generated language, it risks becoming a perfect replica of what investors have already seen that morning.
What changes for founders and VCs when AI writes the first draft
The immediate impact is behavioural. Founders send more outreach because they can. VCs receive more inbound because founders can. So investors develop faster heuristics: they scan for specificity, for numbers that feel real, for customer language that sounds like it came from interviews rather than a model. They also look for signs of genuine insight, the kind that usually shows up as an opinionated trade-off, a surprising learning, or a hard constraint.
At the same time, founders are learning that the deck is only one layer of persuasion. Founder Institute’s programme structure is a good example of how the ecosystem is responding. It does not just ask for a deck. It asks for validation, customer development, pricing tests, a revenue model, and milestone presentations rated on presentation skill and research thoroughness. That is effectively an institutional acknowledgement that storytelling without evidence is weak, and evidence without storytelling is forgettable.
For VCs, the shift is equally practical. If AI makes it easier to produce plausible narratives, then diligence has to go deeper. Investors will lean more heavily on reference checks, customer calls, product usage, and founder interactions. And they will pay closer attention to how founders speak when they are not reading. A founder who can only perform when the words are on a slide is a risk. A founder who can think out loud, adapt, and stay coherent under pressure is a different proposition.
None of this is exactly groundbreaking, but AI is forcing the issue. It is compressing the time between “idea” and “presentable pitch”, which means investors have to be more sceptical about what “presentable” actually signifies.
How to use AI without sounding like an AI-generated pitch deck
The most effective approach in 2026 is to treat AI as a drafting assistant, not a conviction engine. Millard’s critique is essentially about emotional signal. Investors do not need more polished adjectives. They need evidence of insight and the founder’s ability to carry the narrative in person. That means founders should deliberately inject what AI cannot easily produce: the messy specifics of customer discovery, the constraints that shaped the product, and the trade-offs that reveal judgement.
Figma’s guidance offers a practical structure for doing this. A deck should include a problem statement, product overview, business model, market overview, traction and projections, team, funding request, and call to action. The temptation is to fill each slide with generic claims. The better move is to use each slide to answer one hard question in a way that is difficult to copy. For example, a problem slide becomes stronger when it includes the exact phrasing customers use, not just a market-sized pain point. A traction slide becomes stronger when it explains what changed when the founder iterated, not just the current metric (and if there is no metric yet, the founder should say so and show what has been validated instead).
Founder Institute’s emphasis on AI toolkits and AI agents also points to a more mature workflow. AI can help founders organise interview notes, generate alternative slide structures, or stress-test messaging for clarity. But the founder still needs to decide what they believe, what they are willing to be wrong about, and what they will do next if the plan fails. That is the human layer. And it is the layer VCs are increasingly selecting for.
Actionably, founders can pressure-test their decks with a simple rule: if a competitor could swap their logo onto the deck and nothing would break, the deck is too generic. AI tends to produce that kind of interchangeable language unless it is fed strong inputs. The fix is not to ban AI. It is to feed it real material and then rewrite the output in the founder’s own voice.
A brief history of pitch persuasion: from slideware to spoken certainty
Pitch decks have always been a compromise between narrative and analysis. They are short by design, and they are often consumed quickly. That is why templates became popular in the first place. They reduce cognitive load for the investor and force the founder to prioritise. But the best pitches have never been purely about slides. They are about the founder’s ability to make a case, handle objections, and show command of the domain.
The examples highlighted by Figma reinforce this. Airbnb’s early pitch succeeds by focusing on a small number of clear traveller pain points, then backing the story with credibility signals such as press mentions and user experiences. Figma’s pre-seed pitch succeeds by articulating a clear set of principles, access, community, education, that differentiate the product before it exists. These are not “AI tricks”. They are strategic choices about what to emphasise and what to ignore.
What changes in 2026 is the speed at which founders can produce something that looks like a good pitch. That raises the baseline. Investors are no longer impressed by clean slides and fluent prose. They assume it is generated. The new differentiator is coherence across formats: the deck, the email, the meeting, the follow-up, and the founder’s ability to speak without a script. Millard’s focus on oracy fits neatly into that reality. When the written word becomes commoditised, the spoken word becomes more valuable.
The new fundraising craft: combining AI efficiency with human credibility
The practical takeaway from this mini-wave of commentary is not that founders should stop using AI. It is that they should stop letting AI flatten their story. The AI-generated pitch deck is useful for structure, for speed, and for iteration. But it is a poor substitute for the founder’s own conviction, and it can actively harm outreach when it reads as generic.
Programmes like Founder Institute are effectively acknowledging that founders need both: AI-enabled productivity and human performance. Their FI Core schedule explicitly blends AI toolkit setup with deliverables that demand real-world validation and live presentation. That combination is telling. It suggests the ecosystem is moving towards a model where AI handles the repetitive work, and founders are judged more harshly on judgement, communication, and evidence.
And VCs, for their part, are likely to keep tuning out anything that feels mass-produced. Not because they are anti-technology, but because attention is finite. In that environment, the founders who win are the ones who can use AI to get to a sharper truth faster, then show up and speak like they mean it. Simple. Hard. And, in 2026, a big deal.
For founders preparing to raise in the next quarter, the message is clear: use the templates, use the tools, even use the agents. But do not outsource the part that makes an investor lean forward in their chair. That part still has to come from a person.
