Back
EP 004
25 Min
Most vendors sell AI as a way to make a leader look smart and save money. Tamara Wall, who has led support through a 55-point swing in team engagement, says it does the opposite: it magnifies whatever your culture already is.
Tamara Wall does not see AI as software you buy. She sees it as an employee you hire, one that needs onboarding, coaching, and accountability. As Head of Support at Common Room, with prior leadership at Culture Amp and Datto, she has a sharp warning for leaders rushing to flip the switch. On this episode of the Fini Podcast, she explained why AI exposes bad leadership, how to bring a team along instead of scaring them, and which metrics actually tell you the truth.
Meet Tamara Wall
Tamara is Head of Support at Common Room and has led operations at Culture Amp and Datto. She has guided teams through dramatic change, including a period where engagement swung by 55 points, and she is known for building high-engagement teams. Her core belief: support is a critical, strategic function, not a cost center, and AI should make that more true, not less.
AI is a mirror
People talk about AI as a magic fix for broken cultures or weak leadership. Tamara's view is blunt: AI is a mirror. Whatever you already have, good or bad, it magnifies. If you are unclear, inconsistent, or operating without a vision, AI will expose it faster. But if your culture values accountability, transparency, and curiosity, AI becomes a multiplier instead of a replacement. The technology does not set the outcome. Your operation does.
Treat AI like an employee you hire
The most useful reframe in the conversation: introduce AI the way you would onboard a new hire. It needs training data, coaching, performance metrics, and accountability. You would not give a new agent one day of training and expect perfection, so do not expect it from AI either. And when the bot makes a mistake, Tamara does not call it the AI's fault. She treats it as an operational process that failed and needs adjusting. That shift, from blaming the tool to fixing the process, is what separates teams that improve from teams that give up.
The human role evolves, it does not shrink
A common fear is that if AI takes 60 to 70% of tickets, humans are left with nothing but angry, complex cases all day. Tamara reframes it: the human role evolves rather than disappears. When AI absorbs high-volume, low-complexity work, people are freed to solve harder problems, partner with product and engineering, protect revenue, and build real customer relationships. Support becomes a craft you teach deliberately through simulation, shadowing, and new AI-adjacent roles like prompt design and knowledge base management, not something people merely survive.
Her red lines, and the seduction of speed
Tamara is clear about what she will not automate: churn-risk escalations, emotionally charged situations, sensitive identity or safety issues, and contractual or legal commitments. Those require emotional intelligence, and automating them to save time replaces exactly the thing the moment needs. Her warning is that speed is seductive. When volume explodes, leaders panic and chase deflection, and inaccurate AI erodes trust fast. Her system is to pause, look at the actual quality, and pull AI back when the focus should be rebuilding trust rather than moving fast. Move fast, but with guardrails.
What support leaders should take from this
Fix the process, not the blame. When the bot fails, treat it as a broken operational step to adjust, the same way you would coach a person.
Onboard AI like a hire. Training, coaching, quality checks, and accountability apply to a bot just as much as to an agent.
Sell safety, not speed, to your team. People fear losing control and relevance. Show the roadmap, include them in the build, and make them co-owners of the future state.
Set red lines before you launch. Keep humans on churn risk, emotional, legal, and safety-sensitive cases. Those need a human heart and brain.
Drop CSAT for effort and health. Track customer effort, customer health, and product-quality signals. Ticket volume without context is just noise.
Start small and adapt in public. Pilot safely, measure relentlessly, and let your team see you adjust. Waiting for the risk to vanish is how you get left behind.
Listen to the full episode
Tamara goes deeper on leading teams through change, designing escalations, and why CSAT belongs to a different era, in the full episode of the Fini Podcast. You can follow her work on LinkedIn.
AI you can coach, measure, and hold accountable like a real team member is what Fini is built for. Book a demo to see it in action.
Leo: Welcome back. I'm your host Leo from the Fini Podcast. My guest today believes we are all looking at it wrong. She doesn't view AI as software you buy. She views it as an employee you hire, one that needs management, accountability, and coaching. Her name is Tamara Wall. She's the Head of Support at Common Room and has led operations at Culture Amp and Datto. She told me something recently that stuck with me: AI exposes bad leadership. If your process is broken, AI won't fix it, it will just break it faster. Tamara, welcome to the show.
Tamara Wall: Hi, thank you so much for having me.
Leo: Let's start right there. You said AI exposes bad leadership, and most vendors pitch AI as a way to make leaders look smart by saving money. What's the specific mistake you see bad leaders make when they flip the switch?
Tamara Wall: People often talk about AI like it's this magic thing that's going to repair broken cultures or replace weak leadership, and it's really not. If anything, AI is a mirror. Whatever you already have, good or bad, it's going to expose it. If you're unclear, inconsistent, or operating without a vision, that's all going to be magnified. It also depends on where the leader sits. If you're in support, you understand there's more than just implementing AI. If you're a CEO or a finance leader, you just want to be efficient and serve your customer base quickly. What's missing in that is the human loop. If your culture values accountability, transparency, and curiosity, then AI becomes a multiplier, not a replacement, and a lot of people miss that.
Leo: You have a reputation for building high-engagement teams, but the narrative in support is that AI is coming for jobs, especially tier-one. When you roll out an AI tool, what do you say to your team on day one to get buy-in instead of fear?
Tamara Wall: I've led teams through dramatic changes, starting with COVID, and at one point I saw engagement swing by 55 points. The truth is people don't fear AI, they fear losing control, losing relevance, and losing value, especially in support, where people often feel forgotten and less strategic than customer success or implementation. As a leader, my job is to create safety and agency, and I promote psychological safety, because underneath those fears is a concern about being able to provide for their family. I also over-communicate the why. I show them the roadmap and include them in the build process. When your team feels like a co-owner of the future state, not a victim of it, engagement follows.
Leo: How do you maintain that sense of relevance for support teams as AI becomes more integrated?
Tamara Wall: The biggest thing is introducing AI as an employee I hired, no different from interviewing and bringing someone onto the team. When you present it as a partner rather than a tool that just cuts tickets, reception is better. AI isn't only about deflection. I've worked with many neurodivergent people, and I've encouraged them to use AI to tailor their tone to what a customer needs. Presented as someone to come alongside you and help, I've seen better engagement.
Leo: If it's just another employee, how do you performance-manage it, and what does accountability look like when the bot messes up?
Tamara Wall: Because I treat it like an employee, it needs onboarding, you train the data, it needs coaching, and it needs performance metrics and accountability. I check the quality of my bot the same way I'd check quality for a support engineer: how are you responding, what experience are you giving, do customers know they're talking to a bot? I wouldn't give a person one day of training and expect perfection, and I don't expect that from AI either. When AI makes a mistake, I don't deem it AI's fault. It means an operational process failed, so I look at what broke and what we need to adjust.
Leo: If AI handles 60 or 70% of tickets, humans are left with the complex, angry, high-stakes problems all day. Does that make the human job harder, and do we need to rethink compensation?
Tamara Wall: It's a valid concern, but the human role is evolving, not shrinking. When AI absorbs the high-volume, low-complexity queue, it frees humans to solve harder problems and partner more closely with product and engineering. They have critical insights that help shape the product and the go-to-market strategy, and they can protect ARR and revenue risk. They can build real relationships with customers, which matters in SaaS where you talk to the same customers repeatedly, and that feeds upsells, feature opportunities, and voice of the customer. So it elevates support work rather than diminishing it.
Leo: Have you seen AI link support to product in a faster way in practice?
Tamara Wall: Yes. When you're in the thick of tickets you may not see patterns right away. AI identifies them more quickly, so you can improve documentation and give product the insights to make changes. It comes down to trust and quality control: lean on the reporting and speed of AI to surface issues you might miss day to day, so you can make better business decisions.
Leo: You climbed from the phones to the C-suite. If AI takes tier one, what does the new entry-level job look like for someone breaking into CX?
Tamara Wall: It's structured skill development. You shift from learning by osmosis to simulation, shadowing, rotation, and scenario-based learning. You also introduce teams to AI roles, because you still need people for prompt design and to manage the knowledge base that feeds the AI. Tap into people's curiosities and push them toward roles that are still important to support, and use AI-driven practice environments. Support becomes a craft you teach intentionally, not something you survive through.
Leo: You've said AI is terrible at reading renewal risk or sentiment. Is that your red line, page a human with no exception?
Tamara Wall: Some things you automate and some you don't. Speed is seductive, but when you're not accurate or you miss the right signals, it destroys trust. My personal red lines are churn-risk escalations, emotionally charged situations, sensitive identity or safety issues, and contractual or legal commitments. Those require humanity, a human heart and brain. Automating them to serve people faster replaces emotional intelligence. You have to balance the seduction of speed with decisions and telemetry: accuracy, deflection quality, sentiment after AI answers, and how many human touchpoints were avoided without friction. Move fast, but with guardrails.
Leo: What's your system for holding the line on accuracy when the pressure is on to deflect everything?
Tamara Wall: I'm a let's-pause-and-get-the-lay-of-the-land person. I'll hop in myself and look at where we're at, and if speed isn't the direction, I have no problem calling that out and refocusing on quality. It's okay to launch AI, but be mindful of when to pull it back, when efficiency isn't the focus and you need to rebuild trust. Especially in high-volume periods, I check whether we're diminishing quality or moving too fast, and pull back if needed.
Leo: Which applications is AI well suited for now, and which is it not quite ready for?
Tamara Wall: The how-tos, for sure, if it can point to an effective knowledge base, get those out of the queue. I've also used Intercom's Fin to help customers make account changes, or get them to a state where support is just actioning the task. Those are great fits. But when a ticket has been open a long time and the customer is frustrated, that's not for a bot. AI is getting better at appearing human, but it's still missing emotional intelligence, so very low-complexity tasks are great, and the situations needing a human touch are still not there.
Leo: What does a well-designed escalation look like, and what do most companies get wrong?
Tamara Wall: I'm neurodivergent, so I use AI to check that what I'm saying is concise. Where support leaders go wrong is when they no longer have a voice of their own, and it all comes from the AI. When you respond generically, you miss the context, the sentiment, and the critical cues that help you course-correct. A well-run escalation is one where I've gathered context and done my homework, then use AI to proofread or soften the response, not to think for me. And AI doesn't do the follow-through. If something is escalated to me, I own it until it's resolved, regardless of where it broke down. That accountability is mine, not the AI's.
Leo: If you could only have three metrics for your team, which would you use?
Tamara Wall: Customer effort, which is underrated. Customer health. And product-quality signals, like feedback patterns and preventable work. CSAT is antiquated to me, and I wouldn't focus on ticket volume, because volume without context is noisy. What I care about is impact: how much effort it took to work with my team, and how we're protecting and growing ARR, because support is too often seen as a cost center rather than a strategic function.
Leo: Why do you think CSAT is outdated?
Tamara Wall: Because it doesn't capture the true impact or the whole journey. You complete a ticket, send a survey, and get a response for that moment, but support often lacks the health context a CSM or account manager has. You're asking someone to grade an interaction where the agent didn't have the full picture, which isn't a useful way to coach or to get real feedback. CSAT is easy to track, but easy doesn't make it effective. I'd gauge effort instead.
Leo: If a head of CX is holding out on AI because they're afraid of the risks, are they being prudent or making a career-ending mistake?
Tamara Wall: Career-ending is strong, but they need to jump on it. AI is here and our space is already evolving. Start small. It's not all or nothing. Think of it as adding a new workflow, not launching a rocket. Pilot safely, measure relentlessly, and adapt publicly so people see you adjusting. If you wait until the risk disappears, you'll get left behind. The companies winning today are the ones willing to learn in real time.
Leo: Quick fire. True or false, in three years the Contact Us form will be dead.
Tamara Wall: No, but it becomes dynamic and adaptive. It'll meet users where they are, not where your website forces them to go.
Leo: Which is more overrated, CSAT or deflection?
Tamara Wall: Honestly, ticket volume, volume without context is noise. But of those two, CSAT.
Leo: Finish the sentence: the best AI strategy isn't about deflection, it's about...
Tamara Wall: Discipline, making deliberate choices today that set you up for success and create a future your team deserves.
Leo: Tamara, this was a masterclass in leadership. Where's the best place for people to follow your work?
Tamara Wall: LinkedIn is the best place. I love to connect and meet new people, so please send a follow.
Leo: Perfect, Tamara, thank you for joining us. And for everyone listening, if you want to see how the best leaders are navigating the AI shift, make sure you're subscribed to the Fini podcast. We'll see you next time.
Tamara Wall: Thank you.
What does it mean to treat AI like an employee you hire?
Tamara Wall introduces AI the way she would onboard a new team member: it needs training data, coaching, performance metrics, and accountability. You would not expect perfection after one day of training a person, and the same applies to AI. When it makes a mistake, she treats it as an operational process to fix rather than the AI's fault.
Why does Tamara Wall say AI exposes bad leadership?
Because AI is a mirror. It magnifies whatever a culture already has. If a team is unclear, inconsistent, or lacks vision, AI makes that worse and faster. If the culture values accountability and transparency, AI becomes a multiplier rather than a replacement.
Which support metrics matter more than CSAT?
Tamara prioritizes customer effort, customer health, and product-quality signals over CSAT, which she considers antiquated because it captures a single moment rather than the whole journey. She also treats ticket volume without context as noise.
Which support tasks should stay with humans?
Churn-risk escalations, emotionally charged situations, sensitive identity or safety issues, and contractual or legal commitments. These require emotional intelligence, and automating them to save time removes what the moment actually needs.
