With the release of AI agents specific to Financial Services, the recent incidents of export controls on AI models and the subsequent release of such models with safeguards, I have been thinking a lot about how do we maintain our individuality and creativity while embracing an AI-enabled / AI-first approach – in the right (and responsible) way.

The top three characteristics that I feel are necessary for being ‘I, Me, and Myself’ along with AI as a partner are as follows…

This is especially from a Finance / Risk perspective for the near / mid-term while AI becomes more ubiquitous.

Views are personal and do not represent that of any organization…Read on for more and also if you want to understand how all this maps to Family Man Season 3😊.

Context setting:

Cal Newport, in his book ‘Deep Work’, says that in the new economy, only three types of people will have a unique advantage – “those who can work very well and creatively with intelligent machines, those who are best at what they do, and those with access to capital.” These are the top 3 drivers that will ensure someone can excel. While access to capital is a different topic, how does one achieve the first two without the actual experience of deep learning, doing, making mistakes, un-learning, and re-learning?

With AI today, and the advent of Quantum Computing following closely, we need to ponder how to retain and build one’s individuality (read as IP / institutional knowledge). While automation, modelling, and enterprise systems have been there all along, what is rapidly changing now is the degree of autonomy, the ease of access, and the marked shift away from a deterministic and / or classical, reasonably range-bound model-driven explainability – this is on account of the plethora of increasingly efficient LLMs, and also the opportunity to embrace the same for order-of-magnitude gains in productivity.

Approach for Being I, Me, and Myself, with AI:

For managing this transition, I suggest a three-pronged approach, at least for the foreseeable future, as we head to a new normal and continuously adopt and adapt (note that some of the examples here may pertain to the Financial Services, Risk and Regulatory domain).

1. Learn without AI and with AI while embracing AI:

This has been deliberately framed in the order ‘without AI and with AI’. As practitioners, we should learn what the process should look like by actually doing it ground-up first without AI (‘process’ here refers to any business process, workflow, area of work, or any general outcome) – this helps in building foundational understanding and expertise.  Some simple examples here from a Finance and Risk standpoint – does one know how the Excel valuation or risk model works, what happens when one changes the formulae, links, or numbers, how the system in the current state works, how the data flow looks like end-to-end, where the process sits across the Front/Mid/Back office, what are the required business outcomes, and what should be the end state?

Complement this understanding by then learning and visualising the process with AI – understand what the output from AI is, where it works, and where it fails. Importantly, how can you be more efficient – can you adopt an AI-first blue-sky approach by making AI work for you and then correct or fine-tune the output as needed – this will help you be more productive.

Remember, AI can give you the same outputs in seconds that could have taken hours / days to achieve – if the process is re-imagined the right way and the data is correct. To reach there, one should first understand the business process, the underlying data, and what one wants.

2. Review AI:

As a principle, review your AI output very closely and critically. If one is using AI for making business decisions on financial metrics, try asking AI the same question in multiple different ways and check if the model is providing the same / similar output. Ask yourself – can I explain this AI output to the regulator and reconstruct the same graph / business insight if asked – will it stand audit or regulatory scrutiny – whether false +ves and -ves are fine? If the answer is no, how do you safeguard yourself? – what is the margin of error and is the impact acceptable? – what is the moat? 

Said another way, ensure your AI models are meeting all firm-specific policies, industry-specific model and data-related regulations.

3. Remember HOITL (Human Only In The Loop) while embracing HITL / HOTL / HOOTL:

We may have already heard of Human In The Loop (HITL), Human On The Loop (HOTL), and Human Out Of The Loop (HOOTL) as regards the use of AI and agents. All this is great. But – do not forget to document your process without and with AI; do not forget how to run your process end-to-end as known in the pre-AI model implementation world; do not forget what unique value proposition should be part of your process. Retain your personal or organisational IP and institutional memory, especially in this phase as we continue to learn and work with AI, and it becomes more pervasive.  Ask yourself – how will I manage a business process such as financial consolidation or reconciliation or fraud detection / AML / KYC if the AI model / agent is pulled out for any reason or the cost of tokens have hit the roof or the model results are unable to adapt or if a new extension is needed around the process or there will be a wait-time till an enhancement is made available – do I retain the organisational context and process knowledge? In short, be prepared to at least once in a while run your business process without AI till such time you are fully comfortable – i.e., run it occasionally in a ‘Human Only In The Loop (HOITL) mode’, as you would have done before. Note that AI is a tool for outcomes – it is a means to an end, and it is not an end in itself. Establish a clear governance for AI models, buildout, and adoption.

That said, when you adopt AI – adopt it the right way by reimagining your process, controls, limits, and desired outcomes – do not perform just a simple lift-shift of what is there today.

Conclusion:

I believe that by adopting the above approach, one can build expertise, thrive with AI, and derive the maximum value while embracing AI-led transformation in the right way.

The cornerstone is to be clear about what we want out of AI and to acknowledge the context, culture, and clarity of purpose – around process, data, controls, and outcomes.

On a lighter note, this line from the recent espionage web-series, ‘Family Man Season 3’ summarises the approach well – “Aukaat badhaani chahiye apni…lekin bhoolni nahi chahiye” (You should increase your status… but you shouldn’t forget it). 

Please run a prompt on AI for more context if needed…or…simply just watch the video clip here for fun😊 – it will help us stay grounded.

Read the Disclaimer here.

Credits for video link used above: Amazon Prime