Mike Gualtieri and Boris Evelson of Forrester recently published a great new paper Introducing AI-Powered, Human-Controlled Digital Decisioning Platforms (subscription or payment required) and you should get access to this paper and read it now. It’s got some great content and recommendations and is well worth your time and money. It follows on from previous research into digital decisioning and focuses more explicitly on platforms that support digital decisioning:
Digital decisioning platforms (DDPs) allow application development and delivery (AD&D) pros to combine the best of human decision logic with the best of AI to implement application-embedded automated decisions, i.e., digital decisions. DDPs accomplish this by providing a suite of capabilities that enable business subject-matter experts to define decision logic, incorporate data-driven decision intelligence technologies such as machine learning (ML), govern change, and deploy digital decisions within business applications.
Digital Decisions Are Fast And Frequent Automated Decisions
It’s really important to think of decisions as distinct things. As one of the authors said in a note to me:
a decision [process] is different from a business process
Conflating BPM and Digital Decisioning will not help you succeed with either or with the digital transformations they make possible. In particular, considering decisions as separate things to be improved let’s you think about how you can improve the whole portfolio of applications you already have, not just the business processes you have under management.
The paper focuses on fast, frequent operational decisions. Neil Raden and I introduced the basic classification of decisions used here in our book, Smart (Enough) Systems, back in 2007:
- Strategic, one-time one-off decisions typically made with plenty of time for analysis
- Tactical decisions that repeat, relate to how you run the business and still have some time for analysis
- Operational decisions, about a single transaction that must be taken quickly – often in real-time (and sometimes in “really real-time” for sub-second response).
Digital decisioning focuses on these operational decisions – decisions about a single transaction or customer such as how to handle this claim, whether to originate this loan or pay this credit card transaction, what offer to make this customer etc. Most organizations have many such decisions they could usefully digitize but many struggle to identify them (get in touch if you are interested in our remote decision discovery and decision value assessment workshops). Establishing that decisions can and should be digitized is the necessary first step and #1 of our three critical success factors for digital decisioning – put DecisionsFirst™.
But how best to automate these decisions?
Digital Decisioning Platforms Combine The Best Of Humans And AI
Our #2 critical success factor for digital decisioning is to mix and match technology – find the right technology match for each decision. Sometimes a decision can be automated just with policy or regulatory business rules. Sometimes you need to do some basic analytics to find the right thresholds. Sometimes a powerful mix of business rules, machine learning and AI is required. Don’t get hung up on just one of these technologies, keep the whole range of technology in mind as you automate your decisions. The paper has some great discussion of this critical point to which I would add a couple of observations from our work with clients around the world:
- Use decision models to understand your decisioning problem and find the right technologies to automate it. It’s hard to pick the right approach to a decision that’s just described in a requirements document. Build a decision model using the Decision Model and Notation standard first.
- Don’t code decision logic using programming languages or scripting languages. Just don’t. It’s too hard to change, too slow to change, too expensive to change and lacks transparency. Find a better way – use a Business Rules Management System to manage this logic declaratively, use decision modeling to maximize the value you get from a BRMS, and write these rules to engage the business – even if IT still changes them! Digital decisioning requires agility and transparency and only a ruthless focus on managing your decision logic will deliver it.
- Stop separating your operational systems from your analytic systems. Treating analytic systems as something distinct from operational systems reduces the value you get from your data and prevents effective data-driven decision-making at the front line. Digital decisioning using rules and analytics lets you deliver more value from your data by pushing analytic, data-driven decisioning to the front line of your organization. We often walk clients up a simple analytic sophistication curve:
- Model an operational decision and automate it using business rules based on policies, regulations and best practices
- Use data to validate those rules, check thresholds and clarify impact e.g. does that threshold really identify the top 5% of your customers?
- Apply simple descriptive analytics to identify means, standard deviations and trends that you can encode in your rules. E.g. take the mean and standard deviation of length of stay by treatment and write rules to flag those greater than one standard deviation above the mean for wastage review.
- Use data mining techniques to classify and categorize your customers and transactions. Write different rules for each category. E.g. What are the different kinds of customers we have based on the channels they use? Does that change the offers we make?
- Identify the predictions that would change and improve your decision-making. Apply predictive analytic, machine learning and AI techniques to build these predictions and then operationalize them in the decision by wrapping them with new rules. E.g. if we can predict who’s a churn risk 60 days before they cancel their service, we could change the logic for the outbound marketing decision to focus more on retention.
One of the key aspects of the paper is the emphasis on human control and the authors make it clear that using the range of digital decisioning capabilities out there does not mean giving up human control. In fact, using these technologies to automate the base decision can improve human control. Freeing humans form making the transactional decisions themselves allows them to focus on how such decisions are made and should be made moving forward. Being able to step back from the day-to-day in this way can really improve your ability to deliver continuous improvement and true transparency.
And that brings up our #3 Critical Success Factor. Always institute a process of continuous improvement – analyze your decision results so you can systematically improve your decision-making. We always build this kind of continuous improvement loop into the solutions we build for clients – not just to ensure human control of automated decision-making but also to create the resilience they need to cope with an ever changing world. Just because your decision is automated with sophisticated AI or ML algorithms, does not mean you have to give up human control. And you don’t need to let your users override your system either – done right (the way we do it), digital decisioning delivers automated responses and human control.
As the authors say:
Decide To Decide Digitally
And if you need help doing it, contact us. It’s what we do.