Six strategies for implementing profitable AI (with little experience!)By Tom Chiles on October 30, 2023
Thanks to the magic of generative AI and large language models, AI hype is in full swing.
Although it’s easy to find articles making it all sound easy, implementing real use cases in business that deliver positive return on investment requires careful consideration of a number of factors.
Machine learning, robotic process automation, process and task mining, business process management — these all sound like tools to help with your operational efficiency problems, but how can you implement these effectively with little experience?
Over the years, I’ve worked with many brilliant companies of all shapes and sizes that have successfully implemented automation and AI at the heart of their business operations. They’ve learnt the hard way the positives and negatives of different approaches to AI implementation — so here’s my attempt to boil it down to a few approaches to consider when starting your AI journey as a newbie.
1. The centre of excellence approach
This involves employing a dedicated team to manage the business change, gather use cases, implement the AI and support it along the way.
I consider this one a heavy lift to begin with; don’t expect any positive return on investment in the short term, and it’s crucial your execs understand this is a strategic initiative for long term benefit — not a tactical answer to your immediate inefficiencies.
This typically requires a large initial upfront buy-in and investment, employing a Head of Automation/AI to build out a team, and will likely require partner implementation skillsets initially dependent on the product you choose, and your appetite for getting runs on the board. Be wary of which tool you choose; more brittle automation/AI tools will require a dedicated support process and team.
Summary: High upfront effort, low risk, low initial ROI — but high long-term business impact if done well
2. The business user enablement approach
This involves giving people who actually make the business decisions the tools to automate them. Depending on the specific business, I’ve seen a range of scenarios play out, ranging from ‘scales really well with the right guardrails’ to ‘we’ve created the wild west.’
The key thing to understand here is you’re asking non-technical people to implement a new tool they don’t understand. The internal business change required to do this well involves new department goals, careful positioning of automation and AI and clear boundaries set and guardrails in-place. Letting loose without people, processes and tooling there to ensure some form of consistency and quality in what is being done can lead to the aforementioned ‘wild west.’
Summary: Medium upfront effort, high risk, high ROI if done well
3. The service-delivered approach for business user outcomes
Implement the ‘hard bits’ quickly with a team of experts. This involves engaging a company that offers both the automation/AI product, and experts to support rapid implementation.
Likely during the RFI/RFP section of your evaluation of automation and AI products, you’ll start to realize that to get short-term value, you may want the company you’re considering to bring in their experts to implement AI for you. The success of this is really dependent on how well the company offers a ‘service wrapper’ around its product, and can involve forking out a significant professional services bill through a long engagement.
Summary: Low upfront effort, low risk, high short-term business impact — but high cost
4. Off-the-shelf AI point solutions
These solutions aim to apply AI to generic use cases without any tailoring to the needs of a particular business. This one can be appealing for organizations newer to AI, and comes with little real risk to the business initially. There are some practical limitations with this approach which limit the power of the use cases they can tackle, particularly the ability to forecast or optimize (dependant on the AI technique applied) without meaningfully understanding the business’ specific processes and guardrails.
Due to the lack of precision, even though these applications can be tempting based on the price, you may not be gaining any operational improvements or return on investment, and they may quickly be part of your software application kit graveyard.
Summary: Low upfront effort, low risk, low short-term business impact — but low cost
5. Modernization of core business systems
This refers to implementing a large modernization program of crucial systems (e.g. ERP, supply chain systems) to their most recent product release. The majority of core business systems vendors now embed some form of AI in their products within their standard set of features, and these are accessible within the applications themselves.
This one is a bit of a rabbit hole; large scale modernization plans typically have very high and continued costs, a long and rarely-to-plan timescale and often open up organizations to risk of business-as-usual disruption. Second to this, my caution here would be around the AI functionality itself — these companies are not AI-first organizations, they’re only adding some subsets of features, so therefore don’t expect the same level of expertise or game-changing use cases offered by AI-first companies.
Summary: High upfront effort, high risk, low short-term business impact — but high cost
6. AI applications for high-value use cases
Here is the sweet spot in my humble opinion: choosing a new tool focused on automating your high value, highly-specific use cases out of the box. These tools are business-user focused, tend to also be industry specific and aim to apply AI that aligns to your current business process and guardrails.
Products like this work incredibly well applying machine learning to specific use cases based on your data, and require little support or continued maintenance effort.
Summary: low upfront effort, low risk, high short and long-term ROI
My recommendation for implementing profitable AI…
Option six, with the out of the box benefits and the ability to customize the AI under the hood if you want, is the perfect middle ground. This is Peak’s speciality, automating your specific use cases within inventory management, customer segmentation and pricing intelligence with your business-specific AI. Whatever you choose, choose wisely!