40% of S&P 500 Companies gave AI prominence in earnings calls over the last quarter.  

"Less than one in six — 16% — mentioned it in their corresponding regulatory filings, highlighting how AI has yet to make a material impact for the vast majority of companies. “The joke out there was that all you had to do last quarter was say ‘AI’ and your stock would pop immediately,” said Bryant VanCronkhite, a senior portfolio manager at Allspring Global Investments, the $550bn asset manager."

Financial Times 22nd August 2023

Does the launch of Enterprise ChatGPT change things by giving enterprises a safe, secure, and practical way to leverage GenerativeAI?

Chris Surdak points out that the Terms of Usage in OpenAI's Terms of use have not changed.

" While eChatGPT might not "learn" off of your data, that doesn't mean it doesn't still collect and have unlimited license to "your" data.

As I heard someone say the other day, "Your data isn't in the cloud, it's on someone else's computer."

I suggest that anyone who cares about #informationgovernance, to say nothing of #informationsoverignty in those 80% of Fortune 500 companies that you mentioned, might want to take a much deeper look at the precise wording in those Terms of Use". 

Surdak has more advice foe enterprises before rushing in where 'angels fear to tread'. That is to be clear what the goal is and not repeat the mistake of the majority of enterprises that have failed to succeed with RPA, automation, and AI projects. 

Effective or Efficient: Choose One-  by Chris Surdak

When deciding how to use ChatGPT it is important to set specific goals for each specific use case. You can attempt to use it for a wide range of reasons across a wide range of use cases. But, in each individual use case, you must choose between improved efficiency and improved effectiveness. Aren’t these the same? Absolutely not. Efficiency means getting the same throughput for less effort (cost). Effectiveness is getting more throughput (revenue, units, etc.). These two goals are not only different, they are generally mutually exclusive. Focus on one or the other, and you will likely succeed. Expect both from a single use case, and you are almost guaranteed to be disappointed.

How can I be so sure of this assertion? I didn’t come up with this rule, W. Edwards Deming, the father of the Quality Revolution did so; 75 years ago. Deming discussed a process of incremental improvement, periodically interrupted by significant process disruption or change. A healthy process cycles periodically between the two approaches. Unhealthy processes try to do both at the same time. In my own experience, Deming was completely right about this rule. Organizational change (technical or otherwise), can improve either efficiency or effectiveness; never both at the same time in the same process.

Why? Mostly because these are polar opposites. Efficiency emphasizes scarcity, Effectiveness emphasizes abundance. Efficiency is zero-sum, and Effectiveness presumes growth. Success in the metrics of Efficiency almost excludes success in Effectiveness metrics. Organizations that pursue both goals simultaneously experience strategic dementia; they lose the ability to think, reason, and remember to such an extent that it interferes with effective operations.

An organization with strategic dementia has difficulty completing familiar tasks. This is exactly what ineffective adoption of automation technologies like RPA or GenerativeAI suffer. If they are trying to automate for efficiency, they want to do the same old thing, only cheaper or faster. Automation can do that. If they are automating for effectiveness they want to do different things. You cannot do both at the same time, hence organizations trying to automate to do both effectiveness and efficiency almost always end up achieving neither.

Aligning With LLMs

The takeaway here is that organizations hoping to effectively leverage LLMs need to first choose those processes or use cases that may lend themselves to automation. Then they must choose, for each instance, whether the goal is efficiency or effectiveness. Pick one, and stick with it, and you will greatly improve your chances of success. This was a key lesson learned over the last decade of attempting to implement RPA, and the difference between success and failure should be even more stark when applied to an even more powerful automation technology like ChatGPT."

Full article here.

Back to hype and reality when it comes to GenerativeAI. 

 “Some companies are saying they’re doing AI when they’re really just trying to figure out the basics of automation. The pretenders will be shown up for that at some point,” said VanCronkhite

On the other side of the coin AI technology vendors are not holding back in their promotion of the powers of AI and, in particular, Generative AI (GenAI). That was ever thus. Gartner places GenAI at the 'Peak of Inflated Expectations' but we need reminding that the next stage is the 'Trough of Dissilutionment'. 

 Therein lies the rub. Stage 3 the 'Slope of Enlightenment' may be expected in the 2024/5 to 2028 timeframe with full deployment over a decade. Those companies that are left behind now potentially face significant competitive threats to survival, never mind growth.

First tip: It is wise to keep one's analog feet on the ground as there are many risks to face in this new world of GenAI and Large Language Models (LLM) before you will reap rewards. This is a time for experimenting as Zurich recently announced it is doing. And GenAI is the tip of a big automation iceberg. 

This experimental stage should be planned with the same care, resourcing and C-Suite support as any new business venture. If your organization has still not solved the challenge of data and system silos inherited over years of M&A what makes you think that GenAI offers transformational opportunity? 

If you have not successfully applied RPA across the enterprise what makes you think that you can leverage GenAI?

What business goals do you want GenAI to solve for you? Here are a few starters. 

  • Reduce the cost of large call centres?
  • Accelerate claims processing times?
  • Reduce fraud impact when writing risk and/or managing claims?
  • Increasing efficiency across complex supply chains?
  • Empowering claims handlers, loss adjusters, engineers, etc. with the rich and complex enterprise knowledge base?
  • Analysing policy and legal documents & wording to ensure compliance, consistency, relevance

Celent's report ( link to it below under 'Further Reading' highlights the potential.

There is a caveat- the potential dangers. All the more reason to choose partners who know what they are doing!

Start with your enterprise priorities and business goals and then let a multi-disciplinary task force research and plan with reliable business and technology partners. I can't help feeling that some of the GenAI  Teams that consultancies have created to 'help' clients are in fact going to use the clients' dollar to learn on the job themselves.

Choose advisors that you can trust. They must have mastered data management and AI in its widest meaning ( GenAI is just one of many tools) and have a track record of delivering outcomes that support business goals and strategies in a variety of industries.

Just recently I listened to the impressive work that Palantir is conducting with existing clients in healthcare, defence, and the public sector for example. Literally life and Death in some cases. Nearer to home Swiss Re's Chief Data Officer Ian Hancock states " We like to describe what we’re doing with Foundry as becoming a data-connected company...This means more innovative solutions and services for our clients, as well as deeper information to steer and drive the business.". 

At another briefing, I heard how Blue Chip organisations such as Liberty Mutual, the FCA, RES, Gigaclear,  and The Department for Business, Energy and Industrial Strategy (BEIS) rely on Aiimi to deliver insights from the complexity of multiple (and incompatible) technology stacks that are similar to those found across carriers, brokers, re-insurers and MGAs.

Take RES, the world's largest independent renewal energy company, whose IT Director  Jonathan Payne states "Having delivered world-leading energy projects for more than 40 years, we understand the value of advanced technology and data capabilities for innovation and continuous improvement. The capabilities and potential of the Aiimi Insight Engine align with our ongoing transformation roadmap, which seeks to uncover and harness the true power and potential of advanced data and cloud technology."

Take time to engage with such specialists to explore the potential and plan to leverage data management and AI for competitive advantage: -

  • Explore your goals and the potential of AI to achieve these
  • Devise data and AI Strategies aligned with company goals
  • Transform the business on strong data and digital foundations 
  • Enable the business with data-driven strategies and self-service access to the benefits of data science, data engineering, and AI.

Aiimi publishes valuable guides in plain English that you may find useful to navigate the transformation of business. Both Aiimi and Palantir explain how to activate LLMs and AI in a private and secure network so that you can plan and experiment safely. 

Palantir AIP

Llama 2: our thoughts on the ground-breaking new large language model.

AI and LLMs are neither new nor a panacea to the challenges of the insurance world. The algorithms and maths of today's AI are similar to those devised in the 1950s. The big difference is the sheer computational power applied to generate outcomes faster. Just make sure they are the outcomes you wish for.

Further Reading

The Insight Opportunity- uncovering the next generation of information intelligence

Navigating AI usage in the insurance sector

ChatGPT Explained: A breakdown of how it works for curious business leaders

Generative AI is Coming For Insurance: Celent Report

By Chris Surdak Part 1: Chatpocalypse Now

                           Part 2- Chat-tastic Results