I want to be precise before I criticize an industry. The legacy speaker bureaus are not bad companies. BigSpeak, Washington Speakers Bureau, Harry Walker Agency, CAA Speakers are real businesses with real talent, built by real people over decades of grinding relationship work. They have agents who know their speakers personally, who have placed thousands of keynotes, and who understand the dynamics of a corporate event in ways that took years to develop.
The problem is not that they do bad work. It is that they were optimized for a world that no longer exists, and that optimization now works against the buyers and speakers they serve.
This is a structural problem, not a small one. It is exactly the kind that opens a window for a different kind of business.
Problem 1: the roster is a time capsule
Walk through the speaker pages of any major bureau and run a simple test. For each speaker listed under "AI" or "Innovation" or "Future of Work," look up their primary credential. When did they actually work inside an AI system? When did they last deploy a model, study a deployment, regulate one, or build a company around one?
In most cases you will find the expertise is in "digital transformation," a term that peaked in relevance around 2018, or in "leadership through change," which is broad enough to mean anything. The AI framing has been bolted on top of an existing career rather than built from inside one.
This is not the speakers' fault. Bureaus select for proven performers, not emerging expertise. The incentive is to book the person who has given this talk two hundred times and will absolutely not embarrass anyone in the room. The incentive is not to find the person with the most relevant knowledge for the problem the organization actually has.
The result is an enterprise buyer paying a serious five-figure fee for a keynote that says things they mostly already know, framed in a way that feels current and is not. The ten kinds of AI speaker guide sets out what the alternative looks like, and the test question for each one.
Problem 2: the diversity gap is structural, not cosmetic
Legacy bureaus will tell you they are committed to diversity. They will show you the speakers of color on their rosters. They will point at a dedicated diversity speaker list.
What they will not tell you is that the structure of the model itself produces homogeneity.
Here is why. Bureau agents work on relationships and build books of business over years. A new speaker who wants representation needs to arrive with a profile already assembled: a book, a TED talk, a media presence, a corporate speaking history. All of those require resources, networks and time, and all of them are considerably easier to accumulate if you started from a position of access.
So the speakers who get representation are disproportionately the ones who already had the relationships to build the profile. The people with the most relevant emerging expertise, Global South founders, immigrant practitioners, researchers from non-elite institutions, people who built real things inside real organizations rather than writing about them, are invisible to the model.
That is a problem of infrastructure rather than intention, which matters, because it means the gap will not close by asking existing bureaus to try harder. It closes when different infrastructure exists. The practical version of that argument, aimed at your next lineup rather than at the industry, is in AI is a global story and most keynote lineups are not.
Problem 3: the economics reward the agent and nobody else
Industry commissions typically run somewhere between 20% and 30%, and in most cases that number is inside the fee you were quoted rather than beside it. You often never learn what the speaker actually received.
For that share, what do you get? An introduction. A contract template. Someone to chase the speaker if they go quiet. Occasionally, help with logistics.
What you do not get: a place where you can search by specific expertise, see verified credentials, compare candidates on fee and availability, or talk to the speaker before you commit. None of that exists in the legacy model. It is still, largely, a phone call business.
The speaker's side is no better. Out of what lands in their account they fund their own marketing, website, coaching and video production. The bureau represents them without investing in them.
| Legacy bureau model | What a marketplace should do | |
|---|---|---|
| Commission | 20–30%, usually folded into the quoted fee | 17.5%, printed as its own line on the quote |
| Discovery | Call an agent, get their three relationships | A written brief, verified names back in 24 hours |
| Verification | Implicit in the agent's personal say-so | Explicit, documented, the same checks for every speaker |
| If the speaker cancels | Negotiated case by case, usually informally | Replacement in 72 hours or you pay nothing |
| Quality risk | Yours | Rated below 8/10 by your audience and the commission comes back |
| Speaker support | Representation | Representation plus the tools to build a practice |
None of the right-hand column is difficult. It is simply not what a phone call business is built to do.
Problem 4: discovery is broken
Here is a test. Go to any major bureau site and try to find a speaker who is specifically expert in AI governance frameworks for healthcare organizations.
What you will find is a search that returns results by broad category, Healthcare or Technology or Innovation, and then a list of speakers whose bios may or may not mention anything relevant. You cannot filter by industry vertical and specific AI topic and fee range and availability window at once. You cannot see what other L&D leaders thought after the event.
So you call an agent. The agent tells you about three speakers they have a good relationship with. Two are slightly wrong for your audience. One is right. You book the right one, and it takes four weeks.
This is 2026. Every other category of B2B procurement moved to structured, searchable, reviewable processes years ago. Speaker booking did not. Our version of the fix is deliberately unglamorous: you send four things about your event and we do the search, which is set out step by step in how to book an AI speaker.
Problem 5: the AI era created expertise nobody has mapped
The most important thing to understand about this moment is that it has produced an entire category of expertise that did not exist five years ago, and the existing speaker infrastructure has not mapped it.
Consider who your organization most needs to hear from right now:
- The CISO who has spent two years deploying AI systems in a regulated environment and knows exactly what breaks
- The organizational psychologist studying how humans collaborate with AI agents in real work settings
- The healthcare AI practitioner who navigated regulatory approval, clinician trust and patient data governance at the same time
- The founder from Nigeria or India or Brazil who built an AI company with no local cloud infrastructure and no venture ecosystem and shipped something remarkable anyway
- The labor economist who can tell you what workforce transitions look like at sector level, rather than the think-tank version
None of these people are on a legacy roster. Not because they are unknown, but because the model was never built to find them.
What the replacement looks like
Replacing this does not require genius. It requires building the infrastructure the industry should have built a decade ago, using the expertise map of 2026 rather than 2016.
That means structured speaker profiles with verified credentials and specific expertise tags. Range built into the sourcing model from day one rather than filed under a separate tab. A commission that reflects the efficiency of software over relationship maintenance. Tools that help a speaker build a practice rather than just take bookings. Evidence as the credibility signal instead of an agent's enthusiasm. And a way in that does not require you to already know somebody.
What we are building, and how far along we are
iShruti is the bureau for the AI era. We started with a deliberate sourcing strategy: find the practitioners, researchers, founders and specialists doing AI transformation now, verify them properly, and make them bookable.
I should be exact about our stage, because a piece criticizing other people's overclaiming has no business overclaiming. We launched in June 2026. We are early. There is no public roster page yet, because we will not list a name we have not finished verifying, and what we verify is published so you can judge whether the bar is high enough.
What is live is the process. You send four things about your event, with no account and no discovery call, and verified names come back within 24 hours. Our commission is 17.5%, printed on the quote next to the speaker fee, and the fee bands themselves are public in the 2026 speaker fee guide. If a speaker cancels, we replace them within 72 hours or you pay nothing. If your audience rates them below 8 out of 10, the commission comes back. Those terms are on the guarantee page in full.
The bureaus that have run this industry for forty years are not going away tomorrow. But the window for a model built for this moment, with this expertise map, for these buyers and these speakers, is open now. If you want to test it against whoever you use today, send us a brief and compare what comes back.