Implementing artificial intelligence successfully

In the previous article, we looked at how AI is enabling organizations to deliver business transformation quickly, incrementally, and effectively, with minimal disruption and a positive impact on the customer/user experience.

Using design thinking to optimize outcomes delivered by AI through “golden paths,” businesses can use a portfolio approach to transformation that will achieve better value. This leads to faster user adoption, the confidence to change and enforce policies to drive fewer exceptions, and enhances the benefits of the transformation.

Customers and users now expect businesses to deliver continuous transformation, using AI to enhance their experiences. Businesses that don’t will become laggards, exploited and cannibalized by new AI-native entrants as well as their more AI-capable established competitors. This was reinforced in the responses to a survey that Capgemini conducted in 2017 with early adopters of AI-first transformation.

The challenges of implementing transformation

In our experience, transformation often fails at the implementation stage due to a lack of alignment between the solution and the business priorities. It is important that the implementation team works with the design team to avoid the following pitfalls:

  • Designing for exceptions – solutions are too often designed to allow for failure. A fallback mechanism is included in case the golden path fails, such as a call center facility as a backup for the website or mobile app. The solution becomes too large and complicated, the implementation becomes time consuming and costly, and the project loses momentum
  • Measuring the wrong things – solutions are prioritized and selected for their impact on efficiency rather than delivering a better customer/user experience. This leads to negative feedback and poor adoption from users leading to new workarounds and patches. The implementation loses business support and fails
  • Internal competition and lack of alignment – compromises are made on the chosen solution and no one takes ownership of the resulting sub-optimal design. The individual components may be excellent, but there is no synergy and the whole is less than the sum of its parts. For example, IT may have a preference for implementing standard out-of-the-box functionality that could conflict with the expectation of the business for a high degree of customization.

To help avoid these challenges, we use an implementation methodology called ESOAR. This helps the implementation team challenge the design team constructively to minimize potential failure.

E is for Eliminate

When implementing an AI-first transformation, you should take the opportunity to identify and eliminate all unnecessary and sub-optimal transactions/interactions. Focus on keeping customers/users on the golden path and refuse to design for any task, activity or service channel that falls outside that ideal.

For example, if you have a golden path for signing up new customers via a mobile app, but customers are still phoning your call center, then you could stop publishing the phone number and/or have a recorded message that points them to the app. Otherwise, you will be using valuable resources to support a sub-optimal service.

When you make your golden paths intuitive, fast, and effective, then all other pathways can be removed. Your customers/users will get the best possible experience and you will save costs. It is likely that more than 80% of your transactions already follow a golden path – but don’t stop there. Improve the experience further, stop the root causes for exceptions and drive compliance to 100%. If you don’t, you risk a competitive decline in your core customer/user satisfaction.

Stopping your transformation teams from planning for exceptions will require clear corporate policies that reinforce the business strategy. Your design and implementation teams can then focus on areas that create business value with high business volumes. As a result, the transformation activity will deliver early benefits to important stakeholders and gain positive momentum.

S is for Standardize

After the elimination phase has removed all the unnecessary and sub-optimal activities, you should be left with a list of “golden paths” for your transactions and interactions. This list must be validated by a broad representation from the business, who should approve standards for customer/user experience, effectiveness (service levels), business value, and efficiency.

These standards create the parameters for the minimum viable outcome of the transformation. The implementation should not proceed unless these key measures of success have been documented, reviewed, and approved.

A set of standardized golden paths with expected outcomes will give clarity and focus to the implementation team. They are able to quantify what success looks like and discourage non-compliance. This will stop you measuring the wrong things, and ensure your solutions are better aligned to business priorities, and lead to good feedback and higher adoption from users. Your implementation will gain support and be viewed as a success.

Standardization is too often associated with poor user outcomes, this is generally a result of setting the wrong standards, rather than a reflection on the effort to standardize.

O is for Optimize

The standards created in the previous phase are used as design criteria to optimize the new operating model. Traditionally, this was done by:

  • Revising the processes
  • To reduce the amount of human activity
  • And supporting this with technology.

AI-first transformation reverses this approach. It uses the “Five Senses of Intelligent Automatiom” framework to:

  • Put technology at the heart of the solution. This recognizes that machines can perform some activities significantly cheaper, quicker, and more effectively than people
  • Re-focus human involvement in areas where they can best impact outcomes and add value. Humans and machines establish a new balance in their respective activities
  • Change underlying processes to support the new operating model.

Businesses are likely to have invested significantly in their existing operating model, and quick wins can be realized by repurposing these assets. Involving the owners of existing assets in the optimization phase can reduce internal competition and create better alignment of purpose. Compromises may be made initially on the implementation, but they must be steps towards the ultimate transformation objective. New investments can be used to improve, augment, and/or replace these assets as appropriate to drive synergies and deliver the minimum viable outcomes.

Aligning all internal stakeholders around the optimized transformation will build strong ownership of the solution and the implementation plan to deliver the outcome. The individual components will work better together to deliver synergies, and the whole will be greater than the sum of the parts.

A is for Automate (intelligently)

As discussed in the infrastructure of AI, the new technology infrastructure landscape is characterized not by monoliths, but by nimble, modular technology. By adopting a “building block” architecture that supports continuous development and dynamic adjustment, resources can be added and removed quickly and affordably.

APIs are used to connect modules that enable the transformation team to continuously update the underlying technology. This may be done by introducing new tools or enabling functionality embedded in existing tools that had not previously been utilized.

The transformation team should plan to prioritize the implementation of tools that add intelligence or new functionality to the solution. It could be a chatbot that gives customers real-time transaction updates, or a data mining tool to augment the expertise of an advisor with real-time insights drawn from a knowledge repository. Customers and users should experience a constantly evolving richness in their interactions.

R is for Robotize

Finally, the team will start to deliver robots. These will not be visible to customers/users, and are simply used to improve the efficiency and effectiveness of the now well-established solution.

The primary benefit of robots is to help free up people’s time from doing repetitive and/or mundane tasks, so that they can focus on adding real value to the new operating model.

Implement change with confidence

AI-first transformation requires strong cooperation between the design and implementation teams. They need methodologies and frameworks that support their work and create a common understanding to enable good communication.

That is why we recommend complementing the “Five Senses of Intelligent Automation” framework for design with the ESOAR methodology for implementation:

  • E – Eliminate all unnecessary and sub-optimal transactions/interactions
  • S – Standardize the golden paths for transactions/interactions
  • O – Optimize the solution using existing investments to drive quick wins
  • A – Automate intelligently to create new AI solutions
  • R – Robotize where appropriate.

ESOAR: Five senses of AI framework

The opportunity to create artificially intelligent solutions to improve customer/user experiences and add value to the business must be seized by the transformation team. Committing to perform each of the ESOAR steps thoroughly and in the correct sequence is critical to maximizing the chances of success.