Recent conversations with customer success and product leaders at the CHURNZERO event in Nashville keep coming back to the same point. The problem we blame on AI is older, quieter, and easier to fix than we think.
A single source of truth for AI means every channel your customers use, from ChatGPT to your help center, academy, and in-product chat, answers from the same governed knowledge. Without it, AI repeats the contradictions between sources faster and more confidently, so fixing sources and ownership comes before scaling answers.
- Why it matters: Customers choose the channel, not you, and each channel is run by a different team with its own tool and content. Weekly releases widen the gap faster than documentation can keep up.
- What the data shows: In a Sage CS Advisory session at the ChurnZero event in Nashville, 78% of CS leaders surveyed use AI after the sale, but only 44% have unified customer data.
- Where to start: Take your ten most-asked customer questions, trace every place an answer lives, and mark where the answers disagree. That list is your exposure and your business case for governance.
Recent conversations with customer success and product leaders at the CHURNZERO event in Nashville keep coming back to the same point. The problem we blame on AI is older, quieter, and easier to fix than we think.
Irana Wasti, CEO of Thought Industries, opened a recent workshop at CHURNZERO in Nashville with a question that made the room shift in its seats:
Have you tested what ChatGPT, Gemini, Claude, or Perplexity tell your customers about your product, and do those answers match your help center, your academy, and your in-product chat?
The reaction was a mix of nods, hesitation, silence, and then? The point of the whole afternoon: “Your customer has so many ways to learn and get help… It’s driven by the customer’s need. It’s not driven by you,” said Irana.
This idea changes how we look at the problem. Customers choose the channel that works for them, not the one you want them to use. They might ask an AI assistant, visit your website, open the help center, or use in-product chat. Each of these is managed by a different team, runs on a different tool, and uses its own content.
The pace makes it worse
Customers expect quick, confident answers because AI tools have taught them to expect that. Each tool has its own chatbot, so mistakes happen faster and sound more certain. Product teams release updates faster than documentation can keep up. Irana often hears from customer-facing leaders who learn about launches after they happen, sometimes with only an hour’s notice.
One conference participant summed it up: it feels like doing a rebrand every week. Each launch touches every customer channel, and most teams don’t have enough people to keep up.
Attendees described the strain in their own companies. One said a technical colleague manages her help center and has never spoken with a customer, only hearing about updates when they go live. Another compared a previous job’s monthly launches, where everyone was aligned and adoption was strong, to today’s weekly releases, which feel rushed and rarely make it to the help center. What’s the right balance?
A customer success manager we spoke with described a quieter version of the same issue. Marketing owns the external materials, so she leaves them alone: “I assume that the people that get paid more than me are doing the right things.”
Has a customer ever gotten a wrong answer? “Probably.”
Any specific example? “Not really.”
The problem isn’t a lack of care. No one is in a position to spot the issue. When a customer finds the mistake first, it reflects on her, and she has to fix it.
A practical 30-day move
Take the ten questions customers ask most. For each, trace every place an answer lives:
Mark where the answers disagree. That list is your exposure, your priority order, and the business case for governance.
The gap between using AI and being ready for it
In a workshop session with Sage CS Advisory at the CHURNZERO event, the same issues came up at almost every stage of the post-sale customer journey: broken handoffs, no single source of truth, and unclear ownership.
One participant said it simply: if the same problem shows up everywhere, “it’s not a workflow issue. It’s an operating model issue.” The facilitators shared an example of a customer with over 30 CRMs after several acquisitions, none fully integrated.
The data from that session showed the same pattern. Of the customer success leaders surveyed, 78% use AI after the sale and 67% use it to scale customer experience, but only 44% have unified customer data, and just 22% use AI for anything predictive. Another finding: 63% of CS leaders have no consistent way to measure AI. When the facilitator asked if anyone measures AI impact across every department, the answer was clear: “No.”
One leader explained why the start matters, for AI and everything else: “If you don’t get off to that good start, you’re dead.”
For AI, that good start means externalizing the right data and clarifying ownership.
Three takeaways
Fix your sources before you try to scale answers.
Every assistant, copilot, or chatbot will repeat what your sources say, just more confidently. Irana put it simply: as soon as you bring together more than one source, you’ll find contradictions. The best thing a team leader can do now is check which channels customers use and what each one says for the same question. Our Thought Industries fix is simple: update an answer once, and it updates everywhere. Everything else follows from that.
Treat knowledge like data: give it an owner and a council.
The group kept asking, “Who owns this?” Answers ranged from everyone to the chief product or customer officer. One company with about 150 people started a biweekly product-led launch sync, with marketing, sales, support, and CS all in the room, a shared Jira board, and a recording for anyone who missed it. After six months, it “seems to be working pretty well.”
Decide what you want the outside world to know about your company, because if you don’t, someone else will answer for you.
Participants noticed that technical help is often locked behind a login, so public AI assistants use marketing pages, which help with sales but not support. Irana shared a story from a large customer whose users got answers on Reddit, where someone said a feature wasn’t their favorite.
“When you’re not in charge of the content that shows up, the content still shows up, whether you like it or not.”
Her advice: make some of your knowledge public, because if you don’t, someone else will answer. Structure what you share so you stay in control.
A last thought
None of this means you should slow down. Irana expects releases to keep coming faster, especially since many updates now apply to agents, not just people. Her advice is to remember that every company has a single source of truth somewhere: “the code has to be written,” so the real answers are in developer docs or product requirements. The job is to pull from those sources.
As one participant said, “I don’t know if there’s a company that’s not” a mess. That’s reassuring, and also a call to act. The companies that get ahead will be the ones whose answers match across every touchpoint and channel.
That’s why Thought Industries launched the Customer Knowledge AI Platform and Lens: to break the knowledge silos that hold businesses and their customers back. Lens connects your sources into one governed knowledge layer, surfaces contradictions before your customers find them, and distributes one trusted answer across every touchpoint, so your customers finally get consistent answers.
Catch contradictions before your customers do
Connect your help center, academy, and docs to Lens and see where your answers disagree. Then deliver one trusted and accurate answer across every channel. Nothing to move, nothing to rebuild.
Start your free Lens instanceFrequently asked questions
Why do AI tools give customers different answers about the same product?
Because each channel, from ChatGPT and Gemini to your website, help center, academy, and in-product chat, is run by a different team on a different tool with its own content. When those sources disagree, AI repeats the contradictions faster and with more confidence.
What is a single source of truth for customer answers?
It is the place where the real answers about your product live. Irana Wasti, CEO of Thought Industries, notes that every company has one somewhere because the code has to be written, so developer docs and product requirements hold the real answers. The job is to pull from those sources.
Who should own customer knowledge?
Treat knowledge like data: give it an owner and a council. One company of about 150 people runs a biweekly product-led launch sync with marketing, sales, support, and CS, a shared Jira board, and a recording for anyone who missed it. After six months, it is working well.
Should companies make product knowledge public for AI tools?
Yes, at least some of it. When technical help sits behind a login, tools like ChatGPT and Perplexity fall back on marketing pages or third-party sites such as Reddit. Deciding what the outside world should know, and structuring it, keeps you in control of the answers.
How do you check whether customer answers match across channels?
Take the ten questions customers ask most and trace every place an answer lives: public AI tools, website, help center, academy, in-product chat, and your support team. Mark where the answers disagree. That list is your exposure, your priority order, and the business case for governance.
Are customer success teams ready for AI?
Most are not yet. In a Sage CS Advisory workshop at the ChurnZero event in Nashville, 78% of CS leaders surveyed use AI after the sale, but only 44% have unified customer data, 22% use AI for anything predictive, and 63% have no consistent way to measure AI.
How does Lens keep customer answers consistent?
Lens, part of the Thought Industries Customer Knowledge AI Platform, connects your sources into a governed AI knowledge layer, surfaces contradictions before your customers find them, and delivers one trusted and accurate answer across every channel. Nothing to move, nothing to rebuild.
