Selecting AI Solutions For Your Customer Journey Should Also Depend On Capabilities Beyond AI

The advent of artificial intelligence has made it possible to meet several moonshot challenges across enterprise functions. From predicting events well before downtime to recommending prices to unblock hard-to-crack consumer groups — the canvas has been ever-increasing over the last decade.

The bulk of enterprise investments in AI, however, continue to focus on customer engagement functions. Spotting opportunities for upselling and improving customer satisfaction and retention continues to drive enterprise AI spend beyond digital show pilots. In addition, the pandemic has highlighted the need for AI intervention in easing customer journeys. For the next few years, we can therefore expect selecting the right AI tool for solving customer journeys to be high on a VP, CX, CDO or CIO’s agenda. 

The technology landscape in customer journeys stayed loyal to ERP and CRM. Over the last decade, this rather sterile landscape has unraveled with the coming of sharply defined point solutions delivered as software as a service (SaaS). What was clearly the domain of monoliths such as SAP, Oracle and later Salesforce now includes specialized best-of-breed tools focusing on personalization, mobility, price tracking, social, feedback capture, response management and sales force productivity, just to name a few.

Best-of-breed tools typically go through multiple trials across divisions before a selection decision is made at an aggregated enterprise level. This has led to enterprises often using multiple SaaS licenses in the same space across different parts of the enterprise. In addition, the speed of globalization since the turn of this century has increased geographic and language footprints across successful enterprises. Last century’s boutique Indonesian noodle maker now sells in Italy, and an Italian coffee machine is jostling for shelf space in Surabaya.

The B2B customer journey map consists of the following:

• Concept search.

• Search for a specific product.

• Product data sheets and other details.

• Comparison with competition.

• Pricing, TCO and business benefit.

• Order, delivery and payment.

• Post-installation or implementation support.

• Warranty issues.

• Downtimes and SOS.

• Feedback.

• Repeat purchase, as well as upgrades.

SaaS add-ons across the customer journey have brought the criticality of enterprise business process management and workflow platforms back to the table. Continuous improvements mean the continuous need for new workflows that improve business processes with new data elements, improved mobility via added channel variants, updated architecture, redesigned governance and changed downstream actions.

If we think of enterprise processes as software programs, then workflows are the controls that enterprises define to ensure adherence to quality, consistency and statutory compliance. These enterprise processes (i.e., programs) work on data using the said controls leveraging native human intelligence of the corporation to make decisions every few minutes. That’s the world we all grew up and lived in until a few things changed for good at the turn of the century.

Distanced operations exploded on a global scale with transaction volumes that enterprise processes and systems were not designed for. Customers started demanding decisions at the speed they were used to thanks to Amazon and Google. Stability and continuity of workforce that created institutional memory was no longer available. New options and bottom-line pressures had created the gigs that everyone started enjoying.

Increasing microsegmentation in an effort to serve new customers as populations came out of last-century poverty, creating pressures on defining rule variants that were no longer humanly manageable. The Coca-Cola advertisement was no longer the same in multilingual and multicultural geographies, and English was no longer enough. With the net and proliferation of mobile devices, data volume and their formats exploded too. There was much talk of data being the new oil. Enterprise processes were suddenly handling multiple types of texts, images and voice all at the same time.

Success in leveraging AI capabilities across customer journeys requires the AI vendor to deliver on 12 critical dimensions to create real impact beyond PoCs and pilots. An effective and successful AI solution should:

1. Integrate with heterogeneous data sources.

2. Read new data without constraints and be able to harmonize across data sources and formats.

3. Extract and understand even implicit intent from input data cutting through mixed languages, acronyms and even emojis.

4. Learn from human actions — recorded from past choices or in real time.

5. Recommend and trigger next-step actions based on AI decisions.

6. Track AI-driven decisions for completion and, in case of errors, seamlessly resort to plan B.

7. Offer abilities to track AI accuracy to allow for process-level adjustments.

8. Adaptation behaviors based on changes in data.  

9. Comply with global privacy norms and regulations.

10. Allow for reconfigurability by enhancing the ease of adding or modifying UX.

11. Make rewriting of an AI workflow a matter of hours and not days.

12. Not make demands on existing business processes to change their underlying data models or systems.

The absence of any one of these capabilities will put the burden back on enterprise IT teams to either devise a method to fill the gaps or enforce restrictions (e.g., no acronyms in input messages, avoid situations where customer PII could be exposed, etc). Such an eventuality will restrict AI coverage and increase the cost of deployment, making naysayers question ROI.

AI deployment necessarily leads to discussions around integration, workflow and business process management. At the end of the day, it’s the ease of plugging in, ease of training (unsupervised, supervised and semi-supervised) and ease of change that would make an AI solution tick.

The success of AI in the enterprise depends on the AI platform’s ability to solve a few core enterprise system challenges beyond what data scientists are hired for.