Automation seems easy, but in reality, it is much easier said than done. Chances are, if you're working in a modern enterprise in 2024, you’ve had a manager or senior executive mention automation or an “AI strategy” to improve efficiency in this new age of optimization. This new era, driven by cutting-edge technologies and data-driven decision-making, is all about finding smarter, faster, and more efficient ways to operate your business. It's a landscape where companies are racing to harness the power of automation to streamline processes, reduce costs, and gain a competitive edge.
Automation Adoption is Hard.
Many times, the incentives for automation are messed up at the start. When an employer wants sweeping automation changes, employees can get defensive, and for fair reasons. Most of the time, automation isn’t directed to alleviate the employees from strenuous or mundane work but rather to help raise the bottom line or meet some KPIs.
This misalignment of incentives makes it tough to automate various organizational actions as people try to fit the solution of automation and AI to a problem. This type of problem-solving is rampant in the consulting world, usually starting with the phrase “How can we…”
“How can we improve drive-thru throughput by 30% at peak hours” doesn't drill into what a potential problem could be but rather invites solutions that hopefully solve the problem without actually defining it. The current discussions around automation fall into this category.
There are really hard issues with automation adoption. First, most of the current automation solutions are just a system integration play. Unless you own an end-to-end environment, you will mostly be stitching together these environments and automating their communication.
Another significant difficulty is, of course, the CapEx needed for automation. It’s a significant investment that you must be able not only to recoup but also ensure that there’s an actual benefit when automation is deployed. This last bit is also critical, factoring in the opportunity cost of automation. Just because there’s a positive projected ROI in terms of labor savings, quality, or productivity does not immediately mean that solution should be implemented. What is the cost of having CapEx spent on that solution vs another potential problem? Just because there is a positive ROI does not mean you should invest, especially if you are running a leaner organization.
But, of course, automation comes with huge advantages. Better data collection on processes, labor redistribution, consistency, increased quality, and other obvious benefits point to the long-term viability of automation in the workplace and the reason for so much attention.
So, how are organizations currently adopting artificial intelligence into their system? If you’re developing an AI product, what is your strategy? Top-down implementation or bottom-up adoption?
Top-Down: Sweeping, slow changes.
Top-down implementation is what you would think of when dealing with large software being implemented in an organization. Usually, upper management or the CIO decides that a new system is needed and pours through months and months of vendor RFPs and has a lengthy 6-24 month implementation timeline that creeps up to 36 months when people actually start to understand how to use the new system. Large-scale implementation is never perfect, but there are some advantages when talking about automation or artificial intelligence (using generative AI as an example).
First, it’s way easier to automate if you can control the environment. By having a closed environment that one vendor, like Microsoft, owns, it’s easier to implement full process automation throughout and implement solutions that cross over information systems and departments.
Some other benefits of top-down include large-scale organizational data collection, more resources available to implement and test, and, of course, depending on the strictness of the organization’s IT teams, a blessing to use the software.
However, these benefits don’t come without apparent downsides:
It’s really hard to implement a new system. It takes a long time, and it’s extremely bureaucratic.
Trying to match a problem to a solution. Most of the time, you’re not solving a problem if you start with automation as the solution. If your organization has bottlenecks in a 3PL system, adding a chatbot isn’t necessarily the answer if poor route optimization is the real issue. Being removed from the actual problem creates scenarios where management wants to implement AI “just to implement AI.”
Sunk costs. If you or I individually use a tool that we found ourselves in and that tool stops becoming valuable for us, we can drop it. In a large organization, a sunk cost fallacy often takes place.
Rigorous user testing and stakeholder management are crucial aspects that make or break a top-down implementation. Get the people who will be positively affected by automation at the table in the beginning.
Bottom-Up: Stitching together a network
To start, what is an example of bottom-up adoption? Slack is a great example. In the early days of Slack, adopting it in your organization wasn’t some huge multi-year contract that each organization signed. Still, in fact, more adoption came from teams or individuals using it at work, and later, after hitting a critical mass of users, the organization could enter a paid contract. Bottom-up deployment is also directed at the people closest to the issues.
Some benefits to bottom up:
Closer to the Job to be Done. Individuals are actually able to solve problems that are close to them, rather than guessing about problems from 35,000 feet.
Faster individual implementation. Usually, you just need to download and sign in!
Customization. With bottom-up implementation, users can customize the automation solution to their workflows, creating a better % ROI.
This bottom-up method doesn't come without downsides, though:
Siloed improvements. Currently, most individual automation solutions are siloed for a tiny part of a departmental workflow, meaning the network effects of true automation are hard to piece together.
Data collection. Another downside here is limited data collection on workflow improvement. It’s easy to see how your personal productivity can be improved, but when it comes to changing an entire workflow, you’ll need a strong level of process insights.
So which solution is right for your product or organization?
Top-Down: If you want to release more AI features into your already established product, this approach is most likely the best. Chances are (if you’re B2B SaaS) you already have an ecosystem product and these new features should be complementary.
If you’re looking to add AI into your organization, focus on a specific problem that is facing one of these problem areas before implementing a solution:
Labor shortages / low bandwidth
Quality errors
Room for productivity growth
This will allow you to have a focused target for implementing automation rather than a blanket LLM strategy with limited effects.
Bottom-Up: If you’re looking to release a new AI product that is initially a point solution, working your way through the employees to land a large contract eventually is probably the better strategy. This way allows you to get closer to your users faster without needing to go through the bureaucracy of system implementation.
If you’re an organization, you probably don’t have to control it much. But encourage employees to see how they can use automation to make their jobs more accessible, and maybe even allocate a small SaaS budget to test out various tools, no questions asked. After a critical mass of a single tool, look for a large enterprise contract with the vendor.
There are a lot of exciting products being worked on, but with product moats and development timelines being shortened and shortened, it’s important to think about your go-to-market strategy and how you can adopt automation in the workplace. This article outlines the different options when adopting AI or pushing a product out to market. Remember to carefully consider your own circumstances before deciding on a strategy.
Looking forward to hitting 2024 with more articles/insights! Please subscribe and share to help expand Prompter’s reach. It just takes a minute!
Cheers,
The Prompter Team