Leading into ASAE Annual, we have listened to hundreds of Association Executives. We hear about the AI fatigue and that vendors all start to sound the same for the first thirty seconds. They'll all say, “expert answer engine.” They'll all have a chat demo. The differences only show up when you ask specific questions - the ones a rushed booth conversation usually skips. Ask the six below at every booth.
KEY TAKEAWAYS
- Every vendor sounds alike in a 30-second pitch. The real differences show up in specific questions, not the demo.
- Betty has 3+ years, nearly 200 associations, ~30 million content items from ~5,000 sources, and 1.1 million+ conversations. Ask any vendor to match that.
- Governance - not features - is the real dividing line: who sees what, what's traceable, what happens when an expert corrects an answer.
- “We're thinking about building our own” is a real option - but only with a clear read on the engineering, infrastructure, and maintenance cost first.
- A platform “connector” bolted onto your AMS or CMS is not a governed knowledge system. Ask where the intelligence actually lives.
- The best AI system compounds. It should get smarter and more valuable every year - not just look good on demo day.
THE SHOW FLOOR
Why Does Every AI Vendor Sound the Same in Indianapolis?
ASAE’s theme this year is Driven by Purpose - fitting for a city built on speed and precision. It’s also good advice for evaluating AI vendors. Betty pioneered the category and have had an AI Assistant available for Associations since 2023. It is this industry leading experience that means we have to share what we know and give back to the industry we know and love. We were driven by purpose in 2023, launched in 2024 and here we are in 2026 still leading the way.
A five-minute demo can make almost any tool look impressive. It can’t show you what happens six months in - when your expert corrects an answer, when your board asks, “why not just use ChatGPT,” when you need a second content source added. That’s what this list is for.
Even the vocabulary has been lapped. “Purpose-built.” “Governance.” “Trust.” Every booth will use them. That’s not a reason to ignore the words - it’s a reason to stop trusting the words and start trusting the answers underneath them.
WHY THIS LIST
Why Is Betty Publishing Its Own Interrogation List?
Because we’d rather you ask hard questions than take our word for it. Here’s what 3+ years gets you:
- 3+ years as a purpose-built AI knowledge system for associations
- Nearly 200 associations live today
- ~30 million content items ingested from ~5,000 sources
- 1.1 million+ member and staff conversations
Ask any vendor on the floor for their numbers. Then compare.
|
|
PIT STOP 1 Is This a System, or a Platform With AI Bolted On? |
Ask: “Is this a purpose-built system, or AI layered on top of your existing product? Assistant, or agent?”
A platform is an infrastructure you build on top of. A system has its own architecture and governs how knowledge moves. “Agent” implies autonomous action; “assistant” means it surfaces knowledge under your control. Vendors will reach for whichever word sounds more impressive - ask them to define it.
- Good sign: they describe a decision made at the architecture level.
- Red flag: they describe a setting you can toggle on - or the good part is on a “coming soon” slide. If the impressive capability is on the roadmap, ask what ships today.
|
|
PIT STOP 2 What Happens to What We Build - If We Leave? |
Ask: “When our experts correct an answer, do we own that permanently - or does it disappear if we cancel?”
A correction should shape thousands of future answers, not vanish into a vendor’s black box. If it’s architectural, it compounds into association-owned knowledge. If it was bolted on, there’s usually nowhere durable for it to live.
- Good sign: they can tell you exactly what you’d lose by leaving.
- Red flag: “not much.” That’s not a selling point - that’s the whole product.
|
|
PIT STOP 3 Run the Governance Audit |
Ask: “Walk me through what happens when a member asks something outside their access tier. Show me the audit trail.”
Generic AI treats all knowledge as equal. It doesn’t know what’s yours, who’s asking, or whether the document it just cited is two years out of date. Run these six checks at every booth:
|
GOVERNANCE QUESTION |
GENERIC AI CHAT |
PLATFORM CONNECTOR |
BETTY |
|
Scope - limited to sources you approve? |
No |
One system |
Yes |
|
Authority - can staff correct what it knows? |
No |
No |
Yes |
|
Permissions - enforced across every source? |
No |
Inherited |
Yes |
|
Correction - expert input becomes owned knowledge? |
No |
No |
Yes |
|
Provenance - every answer traces to a source? |
No |
Sometimes |
Yes |
|
Insight - see what members need, across sources? |
No |
No |
Yes |
- Good sign: a specific mechanism for each row.
- Red flag: “we’re SOC 2 compliant.” Neither is “answers stay within each member’s access level” - that’s usually a permission gate on one experience, not a different experience built for each audience.
|
|
PIT STOP 4 Where Does the Intelligence Actually Live? |
Ask: “Is this native, or a connector bolted onto our AMS or CMS? Whose permissions is it enforcing?”
A connector can look identical to a native system in a demo. The difference: it usually inherits access rules from whatever it’s plugged into, answers from one system at a time, and can’t surface insight across sources.
Some vendors will describe role-based security groups as “built into the architecture.” Ask: whose architecture - the AI’s, or the community platform or AMS underneath it? If you could disable the AI tomorrow and every one of those permissions would still be sitting there untouched, they were never the AI’s to claim.
- Good sign: adding a new content source next year is a configuration change.
- Red flag: it’s a second integration project.
|
|
PIT STOP 5 What Would Building This Ourselves Cost? |
Ask: “If we built this in-house, what would we actually need - and what’s the real time to value?”
“We’re thinking about building our own” is reasonable. It also usually costs more than it looks like:
- Dedicated engineering and ML talent - not just someone who can prototype
- Infrastructure you own end to end: hosting, scaling, uptime, monitoring
- Ongoing model tuning as usage and models shift
- Security, compliance, and audit-readiness, built in
- A permanent roadmap that never really finishes
A pre-built, proven system gets you value in weeks, not quarters - because someone else already found the failure modes first.
- Good sign: the vendor takes the build option seriously.
- Red flag: their whole pitch depends on you never asking this.
|
|
PIT STOP 6 Does It Get Smarter Over Time - and Can You Prove It? |
Ask: “Show me a real customer where this compounded. Not a projection - an outcome.”
The easy question is “which platform already has AI in it?” Almost everyone says yes. The strategic question is “which system should own our knowledge layer?” Fewer vendors can answer that with a number.
Here’s ours: the National Fire Sprinkler Association’s Chat EOD program set a goal of 20,000 questions in year one. It hit 23,500+, across 4+ countries. That’s what compounding looks like, not a projection.
Almost every vendor will now claim some version of “insights.” Ask what that actually means: a dashboard staff can scan for content gaps, or a governed intelligence layer your board can act on. Those are different products that wear the same word.
- Good sign: a specific customer, a specific number.
- Red flag: “our AI is always learning.” That’s not a mechanism.
FIND US
See This at Booth 1225 in Indy
Run this checklist at every booth - including ours. Four minutes, six questions, and a willingness to ask the second one after the first answer sounds good.
At ours, you'll see 3+ years and 1.1 million conversations of answers to back it up.
Book a demo → meetbetty.ai/demo
