Selling Artificial
Intelligence - the gap between POC and reality?
As a great fan
of Isaac Asimov, I have always had a great interest in anything touching
robotics and AI and when one of my former employers asked me for advice on closing an AI deal, I decided to create this blog.
AI is much more than just a technology to help robots and it seems
that AI has countless potential applications in all industries, all around us.
Process improvement
Taking a specific industry like finance, which is one of my favourite markets, AI helps
in many ways:
- Helping to provide faster and better selection for a credit decision.
- Reducing investment risk by analysing portfolio and leveraging risk issues
- Detecting money laundering by spotting suspicious activities and cutting the cost of investigations.
- Providing smart chatbots to help customers and reducing call centres workload.
- Automating these myriads of high-frequency repetitive tasks. Robotic Process Automation (RPA), is of the five emerging technologies that JP Morgan Chase uses to enhance the cash management process.
It is clear that
AI will have a significant impact on reshaping the business environment in the
financial industry.
Can Citizen Data
Scientists Help
However, despite
65% (Forbes) of executives expecting positive change from AI, only 30% of them
have taken steps to implement AI in their companies.
Furthermore, a recent CIO article stated that while most IT leaders are aware of AI's potential, most organizations still lack the ability to effectively tackle adoption.
In my opinion, one way to speed up the democratisation of AI would be, in parallel to appeal to Data Scientist, to approach the “citizen data scientist” and enable them to take their data, run it and apply it against different machine learning and deep learning models…to actually get information out the other end.
Furthermore, a recent CIO article stated that while most IT leaders are aware of AI's potential, most organizations still lack the ability to effectively tackle adoption.
In my opinion, one way to speed up the democratisation of AI would be, in parallel to appeal to Data Scientist, to approach the “citizen data scientist” and enable them to take their data, run it and apply it against different machine learning and deep learning models…to actually get information out the other end.
Data scientists
and their “citizen” counterparts working in tandem in order to get to a quick
proof-of-concept (POC) and get more funding and move forward.
However, it
seems that many of the AI POCs are struggling to transform into production and
real projects.
Improving the chance of POCs success
Having done many
POCs in my career, I could maybe apply some of my experience and discuss what
can be done to improve the rate of transforming AI POCs into real projects.
1- Even though POC are often
carried out on rather simple algorithms using immediately available data, it is
important that all parties consider right from the start the company
environment, existing workflows, security and data privacy challenges.
It is clear that AI is a
disruptive technology and the project will have a considerable impact on the company’s
existing workflows. Ignoring how the AI project will eventually integrate
within the working environment might be a rather costly mistake.
2- In light of the first point,
it would be sensible to select for the first POC a solution that solves a rather
simple process which could limit the working environment challenges and create
a positive momentum within the organisations.
3- For the AI POC to move to
production, the organisation would need the skills to deploy it to scale and
have a structure to support it. Planning and agreeing on a clear RACI chart and
resource plan at the beginning would avoid a potential disheartened rejection
when the management realises too late the cost and effort to deploy and run the
solution.
4- Specifically, for AI project,
it seems that the availability of the data when going live is crucial. Therefore,
working with data scientist to get access to real-world data (as opposed to
manipulated samples) should reduce the gap between “real world” requirements
and POC data set.
5- If, as recommended in phase 1,
we did our homework and understood from the start how the company is going to
deliver the required data, we could work with the IT department to plan the
integration of the solution within the company’s technical environment.
6- For any project to succeed,
you need to have a C-level champion within the organisation who understand the
long-term benefit of the solution and be ready to fight for its survival with
peers.
Conclusion
First, a clear
relation with the business function is necessary to better understand the
business environment and technical challenges to integrating the solution within
the corporate IT environment.
Secondly,
funding of AI projects often come from the business functions budget so having
a champion at C-level enable the project to be real and protected from others
business requirements.
Third, a clear
planning around access to data and integration to other sources enable to scale
the project.
And lastly, given
that AI solutions are often disruptive, read the "challenger sales" book (Matthew Dixon, Brent Adamson) many
times and apply the concepts to your sales process.
----------------- For the people who do not have time to read -------------------
Abstract to the Challenger Sales
- Challenging customers’ thinking. Develop a deep understanding of your customer, and learn to push their thinking.
- Know their market. Teach your customer something new about competing within their market.
- Re-framing. Learn to re-frame the way a customer thinks about your category of solution.
- Control the conversation. Learn to stay in control of the sales conversation, and when to gently apply pressure.
- All products and services. The Challenger Sales Methodology will work for most industries, but for the complex large scale business sale, it’s even more vital