Issue #11: Do We Want A Self-Serve AI Future?

These days, there's no shortage of discussion about whether "AI will eliminate jobs." Just in the past week, NYT's Hard Fork podcast, Kara Swisher and Scott Galloway's Pivot podcast, and even Jon Stewart's The Daily Show dedicated episodes to the topic.

Understandably, the discussion has centered on the considerations for employers and employees, including:

• Near-term, will the automation of repetitive, administrative tasks translate to better jobs (more time for employees to focus on what matters) or fewer of them? Or both?
• Longer-term, will AI result in net job creation or reduction? And is there enough time, investment, and opportunity for workers whose jobs will be meaningfully impacted to develop the skills they need to remain gainfully employed... even if it's in a different job altogether?

What we seem to be spending less time discussing, though, is how an AI-enabled future will affect end users or customers... and whether that's a future they would even want.

For example, in a report by Goldman Sachs last March, the legal industry was estimated to be the 2nd greatest place where work tasks would be affected and jobs would be replaced by AI. While Goldman's analysis was somewhat measured, follow-on articles from various publications and blogs used it as the basis for going the extreme: "Will lawyers be replaced by AI?"

If you've ever needed personal legal representation, you might have been motivated by steep hourly fees to daydream about this possibility. However, I think most clients would be highly reluctant to turn their legal counsel and success entirely over to a chatbot at this point.

Another way to think about full automation is this: AI is enabling a self-service model for the end customer. If it's truly full automation, the AI assistance will allow the end customer to handle it on their own.

With this lens, for anything but the entirely transactional, I'd be hugely reluctant to embrace self-service for any major legal issue as an end customer. Or health/medical diagnosis and treatment. Or buying or selling a home. I can do Google-assisted research with the best of them, but if it's serious enough to consult an expert, personally I want a human-in-the-loop.

That is not to say that AI has no place in the sectors I mentioned. Given the nature of the workflows, outputs, and costs, they're some of the biggest potential beneficiaries of AI assistance and automation, as the Goldman report suggested. However, I'm talking about full replacement of human ownership and guidance, surrendering it to AI. While we are using the worst AI we will ever have again (as Professor Ethan Mollick says, ie it will keep getting better), the distance to fully embracing self-service in these areas feels like a long one, except for the most aggressive DIY'ers.

Another example that's close to home for me is Talent Acquisition, an area I've spent most of the past decade supporting, and which I've been a partner/customer of for most of my career. Within the HR field, TA is widely expected to see meaningful AI impact, given it involves several routine, transactional processes - the kind of things AI typically excels in.

There are certainly real opportunities to automate / assist with TA-related tasks: drafting job descriptions, job ads, and candidate communications; scheduling and re-scheduling; screening and sorting; taking interview notes and helping inform feedback. These are fragments of TA work which, when taken together, can add up to a significant amount of time and resource, across several members of the recruiting team. The payoff of reducing time/energy spent on them while increasing efficacy could be huge.

Do managers really want to self-serve their hiring efforts, with AI and no recruiter to support them? In my experience, the answer is a firm "no." Recruiters serve as guides and experts in the recruiting process and talent market. They help plan and execute, they counsel and cajole, and they provide focus and connect the dots around the hiring effort. They are too often criticized and too rarely recognized.

But they are very much needed, and where organizations don't have a recruiter on staff, they're outsourcing the responsibility or effectively giving it to someone else (often an HRBP or operations person). Technically, even without having a "virtual recruiter" bot, we could give a hiring manager most of the process rights and capabilities that recruiters have: ATS permissions to post roles (w/ automated JD creation) and review / process applicants; natural language search, automated inmails, and prospect tracking in LinkedIn Recruiter; automated scheduling and existing assessment tech / libraries.

However, I'm skeptical that any hiring managers would take us up on the self-serve opportunity. For such a critical outcome - who we hire and our future colleagues - and with a process that can quickly become complex and fraught if managed incorrectly, they want a human recruiter to steer them through it.

There will be instances in which full automation and self-serve is not only possible but also preferred by end customers. You need only look to pre gen AI examples of this today, e.g., the ATM over the teller experience for cash withdrawals, self-serve ticket kiosks over booths at a train station (for most needs).

A few thoughts on customer needs that can preference a human-driven experience over a fully-automated, self-serve one:

1. Personalization: specific circumstances or needs require a specific or even different course of action from what's commonly prescribed
2. Adaptation: the designated system or approach is unexpectedly outdated or ineffectual, due to a change or shock, requiring guidance and an approach that breaks from the norm
3. Judgment: a human judgment call is required, especially where there is not a clear cut answer or it's generally a difficult call to make
4. Significance: the outcome of the work is an extremely important one to the customer, which justifies more expertise and attention
5. Empathy: the topic or circumstances lead the customer to benefit from psychological or emotional support

AI will continue to evolve significantly and (by all indications) quickly, would could serve to change the mindsets and dynamics above, along with the resulting calculus around full automation and self-service. That said, it is important to consider the perspective of customers in the near-term, if you expect them to embrace the changes needed to get you to the long-term.

For now, if you're going down the path of automation via AI, consider whether your customers are ready for a self-serve model for your service.