AI ≠ Automation. Think Augmentation Instead.

Many people think AI equals automation. Instead, we should focus on augmentation. Why?

1. Automation is hard to do well, especially for ambiguous or relationship-focused tasks.
2. Automation prompts leaders to think about reducing costs and headcount, when automation strategies may actually cost more and increase headcount because automation is so hard to do well for certain functions.??
3. Augmentation strategies are far less de-stabilizing to existing companies, and may require less ‘unlearning’ and mindset-change for both employees and customers.

When it comes to generative AI, augmentation strategies are often easier and smarter than attempting full automation.

Think of the science fiction characters from early Star Wars movies—C3P0 and R2D2 were not expected to take over entire responsibilities, unless they were very defined (like piloting in uncrowded spaces, or carrying something heavy). These robots were obviously machines, and were optimized to augment human capability by doing repetitive and/or dangerous work. Later in sci-fi, in Star Trek: The Next Generation, the character Data is a very advanced android nearly obsessed with being and acting human, often with dangerous and/or hilarious results. Why? Because it’s so hard to completely anticipate and train human nuance.

The 80-20 rule

In AI and process work, aiming for an 80% automated and 20% human process can often be more practical and effective than attempting to fully automate a workflow. Fully automating a process requires accounting for every possible scenario and relying on perfect data and algorithms, which can be challenging. By focusing on automating the majority of the process (80%) and leaving room for human intervention (20%), organizations can achieve better results. Humans have critical thinking skills; no matter how much generative AI tools may looks smart, no current AI has anything approaching critical thinking capabilities.

Take the example of T-Mobile —they’re adopting AI tools to help their customer service agents better focus on tricky, complex issues. Customers are beginning to expect that companies are using generative AI tools to get them accurate answers through support channels, but alongside humans who can help interpret edge cases, contradictions or incomplete policies. Meanwhile, some of their competitors are trying to develop completely self-service tools—but still haven't launched.

Help humans be more human—with machines that are ‘kinda human’

Because so many companies jump ahead to the promise of massive cost-cuts, they’re trying to completely automate customer service. Yet, leading chabot company Intercom’s Fin product aims to resolve 50-80% of customer inquiries, not 100%. The #8020Rule fully applies here. Walk, don’t run—reducing human demand by 80% should lower stress on your existing support agents, for example, and allow you to achieve high quality while also learning new technologies. However—don’t lay those agents off! Many of your employees have great skills and could enjoy more of their work without so much pressure to do rote tasks, and the smart ones will contribute to your AI models, as long as they know they have some job security and paths to advancement.

I can’t say this enough—laying off humans because you think you have AI dialed in based on a few cherry-picked examples, a proof of concept or a proposal from an external firm is a recipe for disaster.?

You can’t (just) cost-cut your way to the future.

In general:

1. You cannot cost-cut your way to innovation.
2. You cannot cost-cut your way to great customer service.
3. You cannot cost-cut your way to differentiation—unless you solely compete on price.

Yes, not all of your employees may be necessary in their current roles as more and more AI tools roll out—but let people who want to stay with you stay with you, move to higher-value roles, and/or contribute to your new AI-savvy vision. And integrating new generative AI tools requires a lot of work to tailor and customize them to your tone, ensure ethical use, and interface with your company-specific data and processes. (By the way, those overdue data hygiene and consolidation projects no one wanted to fund in the last decade? Expect to hear a lot about them very soon).

If you automate everything without discernment, you may end up genericizing what is most valuable about your company.

A huge hidden risk of over-reliance on AI, especially with cost-cutting in mind, is a race to the bottom by providing entirely self-service models too soon. In addition to Idiocracy -esque chatbots, over-relying on self-service models erodes trust, limits important feedback loops from your users and clients, and requires escalations to highly-skilled “unicorn”-like employees because you laid off mid-level teams.?

If you look at your objectives and key results, as well as your company’s organizational DNA (or historic ‘special sauce’ way of serving your customers) you’ll find good guidance for what you should and should not attempt to fully automate. T-Mobile prides itself on being the ‘un-carrier’ with a more human touch—so they augmented their teams rather than attempting to replace them. If you’re Google, and your company is focused on self-service, fine. But if you’re an organization like T-Mobile who has tried to distinguish yourself through the quality of your human relationships, the creativity of your design or other hard-to-automate DNA, don’t rush to automate what makes you unique. Instead, look at how AI can you help be even more creative, connected and of service to your customers.

Use AI as an augmentative tool: help your humans to be more human. For example, amongst support teams:

• Allow your teams to focus on solving previously unsolveable problems (such as providing better onboarding support, direct support for customers facing complex challenges, or “edge cases”)
• Increase their facetime with customers (if that's what the customers want)
• Level up skills so agents can do better with VIP clients or complex situations
• Write better documentation
• Engage with product teams
• Create video explainers for common issues (also with AI, like Descript !) or better in-product support.

Nearly every department and business can leverage generative AI (and basic automation) toolsets to allow them to do more of their most valuable, ‘genius-zone’ work. An off-the-shelf solution will not replace humans nor differentiate your firm on their own; off-the-shelf tools are both somewhat generic in nature, and also being applied across nearly every company who’s reading today’s headlines. So to immediately advance with AI, lean in to the human talent in your organization.

Incremental gains can fund exponential futures

The incremental gains you get from generative AI ‘quick wins’ can fund more innovative, exponential possibilities. But only if you reinvest your ‘quick winnings’ into your next generation of operations and offerings. Automation might look like the end goal, but it’s not right for every use case, and it’s rarely easy to do well.

(You can read more about incremental vs. exponential use cases for generative AI in our article on OKRs for Generative AI ).

I’d love to hear more about your own work on both augmentation and automation strategies. What’s working? What’s not? What do you wish you knew back on day one?

We’re launching an AI course on this and many other strategy topics for Generative AI with Tec de Monterrey, beginning June 11—check it out, and if we’ve connected in the past, DM me for an insider’s code.