AI in customer support isn’t a nice-to-have anymore — it’s a board-level priority. But the pressure to "bring in AI" often leads to two common pitfalls: picking the wrong vendor or never getting anything off the ground.
In a recent webinar, Assembled AI product expert Brian Yeh shared an inside look at how companies are navigating AI adoption. From the roadblocks he’s seen firsthand to the tactics that set successful AI implementations apart, here’s a recap of the key takeaways.
AI adoption is moving fast, but AI buying? Not so much. If you're struggling to evaluate vendors, you’re not alone. Brian highlighted four major challenges support teams face when bringing AI into their organization:
Even the most AI-forward companies are figuring it out as they go. Klarna famously cut support jobs in favor of AI — only to walk that back months later when they realized humans were still critical to great service. Meanwhile, OpenAI — the company leading the charge in AI innovation — is actively hiring for customer support roles.
"The reality is, nobody has fully cracked AI in support yet," Brian explained. "Every company has a different approach, and best practices are still being written in real time."
Adding to the complexity, AI procurement doesn’t always sit squarely within support or CX teams anymore. Many organizations are now seeing product teams take ownership of AI strategy, especially when AI-driven automation is a broader company priority.
"I’ve worked with support leaders who were all set to move forward with an AI solution, only to find out that product leadership had already defined the company’s AI roadmap," Brian noted. "If you’re buying AI for support, don’t assume you own the decision — make sure product is in the loop early."
Unlike traditional SaaS, AI pricing is all over the place. Vendors charge per seat, per interaction, or even per word in an AI-generated message. This lack of standardization makes it difficult to compare options fairly.
"I’ve seen pricing models based on per chat, per minute, per email, per API call — it’s a mess," Brian said. "What everybody wants is for value to match what they pay, but AI vendors are still figuring out how to package that value."
That’s why understanding your own support data is crucial before you start evaluating vendors.
"Before you even look at AI vendors, get a handle on your own support volume," Brian advised. "How many calls, chats, and emails are you handling? How long do interactions last? What’s your cost per case? If you don’t know these numbers, you’re flying blind in pricing conversations."
By gathering these internal benchmarks, teams can more effectively compare pricing models and avoid unexpected costs.
Traditional software decisions often come with a built-in safety net — you can pilot a tool, see if it works, and roll it out gradually. AI doesn’t work that way.
"I’ve worked with teams trying to trial three or four AI vendors at once, thinking they’ll compare results and pick the best one," Brian shared. "But in reality, they’re spreading their resources too thin. They don’t set up meaningful workflows, and they don’t get real insights. The result? They learn nothing and end up making no decision at all."
Most AI tools don’t follow the standard per-seat SaaS pricing model. Many include a usage-based component, which means the more AI handles, the more you pay. That unpredictability makes finance teams nervous.
"The irony is, AI is supposed to make support more efficient — yet the way some pricing models work, costs can actually rise if AI does its job well. That’s a hard sell to procurement teams."
Brian has seen AI buying go off the rails in two big ways:
It’s tempting to take the easy route — bundling AI with your existing contact platform. The logic makes sense: fewer contracts, fewer approvals, and maybe even a lower price tag. But there’s a catch.
"Once companies really expand their product set, AI add-ons often become an afterthought," Brian said. "They don’t get the investment they need, and the product ends up feeling bolted on. You might be better off with a best-of-breed approach."
On the flip side, some teams overcomplicate things to the point of paralysis. Long RFPs packed with dozens of requirements make it impossible to move forward.
"I’ve seen RFPs that look like Frankenstein’s monster — every possible AI feature under the sun," Brian said. "That’s how you either pick a jack-of-all-trades, master-of-none vendor — or worse, you never get anything off the ground."
How to avoid these pitfalls:
Even with a strong business case, AI adoption isn’t just a numbers game — it’s an emotional one.
"AI is a sensitive topic," Brian pointed out. "Agents might worry about job security. Leadership might be skeptical. Even if the ROI is solid, you still have to win hearts and minds."
This is why having internal champions is crucial. These are the people who can help socialize AI’s benefits, address concerns, and ensure smoother adoption across teams.
"You need someone who isn’t just pushing AI from the top down but is actively engaging with teams on the ground," Brian said. "Find those champions early, and make sure they have a seat at the table."
Before evaluating vendors, take a step back: Why is your company even considering AI? If your answer is, "because leadership said so," you’re not alone.
Many support leaders aren’t just considering AI — they’re being told to implement it. But without a clear strategy, that urgency can lead to rushed decisions, wasted investments, or AI solutions that don’t actually solve real problems.
"I had a head of CX tell me, ‘My board said if I don’t bring AI in this quarter, I’m in deep trouble,’" Brian shared. "But that’s not a strategy. That’s a mandate. It’s up to you to dig deeper — what’s the actual problem AI needs to solve?"
Most companies fall into one of three categories:
Improving the customer experience. AI should enhance CX, not just cut costs.
"One financial services company I work with added an AI-powered voice agent, and customers actually started requesting to speak with it instead of a human," Brian shared. "That’s when you know AI is working well."
Supercharging agents. AI copilots can boost agent efficiency by suggesting responses, pulling up knowledge articles, and automating repetitive tasks.
"Everyone’s using ChatGPT already," Brian said. "The question is, how do you integrate that into your support team’s workflow in a way that actually helps?"
Scaling without growing headcount. If support volume is increasing but hiring isn’t an option, AI can automate common inquiries across multiple channels.
"If your company is doubling in size, but you can’t double your support team, AI might be your best bet."
Your AI strategy should prioritize one of these goals — no vendor will do everything perfectly.
Once you’ve built a strong case for AI in support, the next challenge is getting from evaluation to implementation without getting stuck in endless decision cycles. AI buying doesn’t follow the traditional procurement mold, which means you need a structured but flexible approach.
"AI buying can’t be a free-for-all, but it also can’t be an overly rigid RFP process," Brian explained. "You need a balance — enough structure to compare vendors effectively, but enough flexibility to adapt as AI evolves."
Here’s a five-step approach to choosing AI that actually works for your support team:
Before you take a single vendor call, get crystal clear on why you need AI. Support teams typically fall into one of three categories:
"If you can’t answer what success looks like in a sentence, you’re not ready to evaluate vendors yet," Brian said.
The last thing you want is to go through months of evaluation, pick your ideal AI solution, and then hit a roadblock with procurement or security.
"It’s way easier to deal with internal friction upfront than to find out later that procurement has an AI policy you didn’t know about," Brian warned.
Who to involve early:
Traditional RFPs weren’t designed for AI. With AI evolving rapidly, rigid checklists and feature comparisons often fall short. Instead, focus on the capabilities that matter most to your team. Here’s a guide on structuring an AI RFP that actually works.
"The worst RFPs are the ones that list every possible feature and create an impossible standard," Brian put it bluntly. "Instead, focus on the three to five capabilities that matter most."
Most teams want to “test” AI before buying — but a proof of concept is more than just a test. It’s an opportunity to get key stakeholders involved early, ensuring AI is evaluated in real workflows, not just in a vacuum. Internal champions can help refine success criteria and build trust with teams that will rely on the tool day-to-day. The more aligned teams are early on, the smoother the rollout will be.
"Sending 100 tickets to three different vendors and seeing who provides the best response? That’s not a real test," Brian explained. "You want to see how the AI actually learns and adapts to your workflows."
One of the biggest reasons AI initiatives stall is that teams treat “buying” and “implementing” as separate conversations. But to get real value from AI, you need to think about rollout from day one.
"The best way to get people excited about AI isn’t another vendor demo — it’s showing them exactly how it will fit into their day-to-day work," Brian said.
The reality is, AI isn’t a plug-and-play solution. Success depends on picking the right vendor, setting clear goals, and ensuring your team is ready to embrace the change.
AI in support isn’t about keeping up with trends — it’s about building a strategy that delivers real impact. Moving too slowly can leave your team behind, but rushing into the wrong AI solution can be just as costly.
As Brian put it: "If AI is on your roadmap, don’t just tick the AI box and move on. Be intentional. Define your North Star, evaluate vendors carefully, and keep a clear focus on business impact."
The right AI investment today can transform your support team — driving faster resolutions, empowering agents, and scaling operations without sacrificing quality.
Want to see how Assembled Assist can help your team scale? Get a demo and explore how AI can drive real efficiency without cutting corners.
