Osman Koç is the CEO of UserGuiding, a platform that helps teams improve user onboarding and build in-app experiences that drive adoption.gettyWith the advancement of conversational AI chatbots and workflow automation capabilities with agentic AI, many product-led teams have heavily turned their attention to customer support and experience automation.AI sales agents, AI customer support agents, email outreach sent via automated workflows … All of these sound pretty efficient, especially for a product-led company, as they work well with other PLG (product-led growth) strategies like automated onboarding, free trials/freemiums and self-serve customer support centers and product academies.However, PLG is more than automating and “optimizing” customer touchpoints. Enhancing your product experience and enabling your (potential) customers to experiment and use your product independently, with little to no “unnecessary” human involvement on your side, is crucial, but you still need to make sure you are there for them and are available for extra support, hand-holding or simply human interaction if they ask for it.Map the user journey for automation opportunities.Automation, AI and self-serve capabilities can genuinely lower friction and often quicken user interactions, as the user doesn't need to wait for your team's working hours to access information. Repetitive and predictable customer interactions that don't really benefit from a human voice, like onboarding sequences, feature announcements, renewal reminders and usage-based nudges, are clear, low-friction moments for automation.In fact, based on what I've seen, users often prefer an automated self-serve experience in these moments. But if you cannot decide what to automate, you need to map the customer journey and detect optimal automation points.To do that, you can start by listing every touchpoint where your team currently interacts with users or where users interact with your product, from the moment someone lands on your website to their first meaningful action inside the app, through to renewal or expansion. Then, for each touchpoint, you should ask three questions: 1. How often does this interaction happen?2. How much does the outcome vary from user to user?3. And how much does getting it (slightly) wrong/generic actually cost?High frequency, low variability, low cost of error: that's your automation sweet spot. Low frequency, high variability, high cost of error: that's where your teams belong.In my experience, there are a few areas where product-led teams frequently succeed with automation and AI adoption. If you're not sure where to start, these are the touchpoints worth looking at first:• Pre-signup and trial activation (AI-powered sales agents/chatbots for plan comparisons, feature availability questions, integration compatibility on pricing pages)• Onboarding sequences and communication (behavior-triggered or persona-based email sequences and in-app flows)• Customer support (AI-powered support agents/chatbots for common queries)• Transactional communication (email, SMS or any communication channel automation for renewal reminders, invoice notifications, feature announcements and the first touch of an upsell sequence)However, I should warn you against making the mistake of automating the entire upsell sequence rather than just the first touch. The moment a user engages, shows genuine interest or starts asking questions, that conversation should move to a real person. Upsell is a high-stakes, high-variability interaction, and it's one of the clearest examples of where over-automating can cost you potential revenue.Identify the specific break points and anxiety spikes.Break points are the moments in your user journey where people stop wanting self-serve and start looking for a real conversation, where saving time is no longer the priority, and they're perfectly willing to wait for an email response, schedule a call or arrange a 1-to-1 meeting, as long as there's an actual human on the other end.These moments tend to cluster around a few common situations, such as high-stakes decisions, unresolved frustrations, complex and/or data-sensitive issues and moments of doubt.The whole point of automating your customer experience and communications is, or should be, to free up your team's time and attention for exactly these cases. Automation handles the repetitive and predictable so that your support, success and sales people can focus on the interactions that actually require them.Additionally, human involvement has become a matter of prestige in customer care.Especially for high-value customers and corporate plans, having a dedicated human point of contact is something people expect and will pay for. Many SaaS companies have started offering personal customer success managers as plan add-ons for exactly this reason. If your higher-tier customers want a real person, giving them one is a signal that you take their business seriously, and nowadays, when most interactions are automated by default, that signal can be useful.Though I must also say that human care and involvement should not just be a corporate or enterprise service. Any user, regardless of their plan size, can hit a moment where they need to feel heard rather than handled by automated AI bots.So, if your team structure and business model allow, you should always leave room for human interaction in your customer service and touchpoints.Remember these key takeaways for a hybrid customer experience.To sum up what I’ve discussed, when mapping your journey for automation opportunities, you can look for signals like:• The interaction is informational or procedural, not relational.• Timing and personalization can be handled reliably by behavioral triggers and alerts.• The cost of a slightly imperfect automated response/messaging is low.Where those signals aren't present, you can think the general rule is to keep a human in the loop.Beyond that, you should also make sure human access is visible and easy to find.One common failure is when users don't even realize a real person is available. A simple, well-placed "talk to a real person" option in the right moment is often enough to change (or prevent) that perception.Finally, in the case of a hybrid customer service/communication and a handoff from AI agents, you should always make sure your team steps in with context. If your automation has already captured session history, onboarding completion or previous interactions with an AI support agent, your team should enter the conversation with that information.​Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?