With the hype around AI, anyone would think it’s all new, and flawless. AI is not new, and unrelenting wave of new solutions is enabled by price-performance of critical infrastructure and advances in supporting technology. Tech vendors slug it out for the hearts and minds (and cheque books) of new customers.
It’s very tempting to look at AI capabilities and see them as a quick fix solution to your business systems problems. But, the reality is that as good as they are, AI technologies are simply tools that can’t be rolled out without considerable forethought and planning.
So, before checking out what AI technologies can help improve your processes, here are seven key points to follow:
- Understand the entry and exit points for the process in detail – how is the process initiated, what information is needed before the process can start?
- Map the Brutal Reality: Why some processes don't work today, where are the tangles, the issues, the non-compliance, and where they breaks down. Don’t worry, very few business process have a clean path.
- Tacit Knowledge Traps: where the user’s “tacit knowledge” drives current manual processes. Manual system break-down over time. A user’s cleaver short cuts become hotspots for non-compliance and errors.
- Manual Data Transfer and Operational Risk: Relying on manual keyboard input for cross-application data exchange—commonly referred to as "swivel-chair integration"—creates critical process bottlenecks and introduces frequent data entry errors. Left unaddressed, these vulnerabilities compromise operational throughput and create significant compliance exposure.
- Monitor process exceptions and omissions: exceptions and omissions are where the real work and value occurs. Spend time to determine how they’re handled (or not). Carefully define how these should be processed in any new solution. This element may feel like the tail wagging the dog but alas, it needs to be nailed.
- Describe the objectives: for the new or amended process. Individuals often dislike change and radical re-designs may not be an option. Gather candid feedback to understand stakeholders’ acceptance of any changes. Keep it realistic and achievable – better to reign early than fix an over complex solution.
- Process Reporting: Real-time and Analysis: Reporting must be viewed from the perspectives of real-time, operational dashboard to aid process management, as well as retrospective in-depth information for compliance and process improvement.
Next steps.
- Discrete Tools or an integrated platform? Consider tools or a platform for the process design work. These can be stand alone or, they can be a component within a Process Orchestration platform. Typically, platforms allow for seamless connections between design and the process building.
- Process Orchestration: these solutions are known as workflow or process orchestration systems. Check out Gartner’s BOAT (“Business Orchestration Automation Technologies”) acronym.
- Comprehensive systems and applications integration – with the exception of legacy systems, there should be no excuses for new processes to automatically verify and update data on existing systems.
The Take Aways.
Deploying technology for the sake of it at problems isn’t fun and it certainly isn’t clever.
Formal process discovery sessions are invaluable and it’s often the exceptions that kill process efficiency.
Prototyping is one way to check the viability of a potential approach.
Automation is typically implemented in small elements during the entire process.
Aiming for 100% automation is a heady target, it’s mostly unrealistic, and rarely achievable. For that reason, provision for Humans in The Loop must be incorporated – these are needed to review, correction, and oversight.
Many solutions will incorporate multiple AI technologies – we’re in the era of Poly Model AI and any platform should offer a wide range of options and be able to clearly measure why one AI Service is better than another in terms of accuracy and price per unit.
.png)
