![](https://34.230.210.75/wp-content/uploads/2021/12/Clouds-1024x684.png)
The last several months after leaving an AI/ML focused company, have been eye-opening in terms of the shift in the conversation about digital and AI transformation. The progression over my career has been steady:
phase 1, scientific computing was a niche requirement to solve specific tasks
phase 2, demonstrating and evangelizing on the part of scientific computing convinced scientific teams that automating analysis and data capture brought value
phase 3, scientists and scientific computing partnered to build the systems to drive an enabling technology platform. (In this phase AI/ML was a conversation on self-driving cars and hype cycles)
phase 4, AI and data are in the driver’s seat
phase 4 feels different, if it can even be classified as a phase. It may be more a series of phases overlapping so rapidly that it is hard to distinguish each separate shift despite that each on its own would have extraordinary impact. In this current phase, technology is no longer a Frankenstein-style add-on to the company in which executives’ eyes glazed over as soon as the word ‘database’ or ‘code’ was mentioned. Data has now become the strategy. It is now driving the visions (and anxieties) of the same executives who were arguing a few years ago about whether Scientific Computing warranted a full-time individual contributor role (never mind a C-suite executive).
What has changed is that while those geeks, who weren’t even looped into the strategy conversation a few years ago, have made profound changes to how business works (and will work). To name just the top few: webservices based computing models have democratized massive compute and storage scalability; mechanisms for collecting data and building previously unfathomable data sets have become increasingly available through digitization, novel web-based technology (social networks, e- commerce), sensor miniaturization, and lab equipment integration; and, advances in AI/ML algorithms coupled with domain understanding have enabled the detection of valuable ‘hidden’ information in the patterns that are beyond human capability to identify.
The impact of what was a curiosity a few years ago, limited to Google, Amazon, and a few academic labs, has gone mainstream faster than any prior shift. As such, we all need to get a lot more comfortable being uncomfortable and embracing ambiguity. In that murkiness of ‘what is next’, one certainty I see is that taking the ‘we’ll get to it someday’ and ‘if it ain’t broke don’t fix it’ approach, that many companies have taken with digital transformation to date, is a fast track to failure and obsolescence. Or put another way, just because your plane is still flying with multiple engine failure doesn’t mean that ‘it ain’t broke’.
Here are some thoughts on AI digital transformation:
The first principle to any AI-digital transformation is to stop treating it like it is a monolithic problem with a single solution and a deployment cycle. It is a fair promise that as soon as your multi-year effort has finally reached its initial recommendations, many/most of the solutions you identified are last year’s news. My philosophy for building company capabilities has always been to think long-term but act short-term. As a manager when I indulged long-term, full-fledge solutions proposed that ran counter to my ‘solve the immediate with an eye to the future’ approach, more often than not we ended up back in the planning stage, reverting to a more realistic solution that both solved current company needs and prepared for later.
The second principle is that AI/Digital transformation is going to take some 3-dimensional (if not 4-dimensional) chess type thinking which can likely only be done by a strong, cross-functional executive (or center-of-excellence) team that has real power and the ear of the CEO. The true dimensions are far beyond the list, but four central ones include:
What does AI enable that you (or your competitors) couldn’t do before?
AI and digitization provide incredible opportunities to break down barriers for new offerings, new ways to engage with clients and partners, gain new insights, and enable new approaches to speed up processes by orders of magnitude. It provides ways to detect and address risk at the same time as it creates new risks (e.g., introducing and entrenching bias). The ‘dark’ flipside of this opportunity is that your competitors (including those you know, those you don’t know, and even some who may not know it themselves) have these tools as well. If they are a startup without years of entrenched process and dogma, they may be better able to capitalize on new technology.
How are your internal processes adapting?
Process improvement, efficiency, and automation are projects that are never done. Complacency that your manual or even old-school digital approach works well enough or (worse) is superior to an unoptimized, but novel technology-driven approach, is detrimental thinking. A company who didn’t automate five years ago, because their manual process was ‘better’, is now unable compete against newer players as demand for their product and the data it produces scales up. Cloud-based scalability and technology flexibility may not be a priority at the moment, but the cost and resources to increase or even maintain capability using on-premises resources leaves your process at risk.
The problem is not limited to scale, but also extends to trapping yourself in a specific technology solution. Not planning for a ‘plug and play’ swap out of modular capabilities will increasingly cause slowdowns. Similarly, undervaluing your data by leaving it in formats with poor accessibility and failing to automate data capture will mean long or impossible lead times for analytics or model-building.
Finally, building the right team is perhaps more important than anything else. This isn’t just a question of hiring data scientists out of top schools—it is far more subtle. Your entire organization needs to think differently about data and its value. This may require shifting budgets, educating your staff, and (more importantly) educating your executives. Equally important is to realize that going it alone is not ‘cool’ or ‘superior’, it is bad business. You need vendors and partners that have a track record of delivering quality work AND adapting to the rapidly changing landscapes AND who get your industry, technologies and supply chains. They need to complement your expertise in the areas that are not your core competency—i.e., so you don’t have to try to beat AI companies at AI.
How does AI affect your customers, partners, vendors?
Do away with the assumption that you understand what your customers and partners need. This goes double for your understanding of what they will need in a year from today. You are rapidly adjusting your business, products and capabilities. Failing to understand that the same applies to clients, partners, and vendors is not a recipe for success. Your clients, partners and vendors are getting smart about AI and data, in many cases faster than you are. Locking clients into underpowered technology products will ultimately drive them away. Their data is exploding; their competition is ramping up; the pace of change is hitting them just as hard as it is hitting you. Is the best course of action pain-staking incremental improvement? Having some corporate ‘empathy’ here is warranted. You also are a customer, client, or partner. Apply the lessons learned from interacting with these companies and how they adapt to AI to reflect on your own company’s products and delivery.
What challenges does AI bring?
The challenges in AI transformation are formidable, but the biggest mistake in addressing them would be to assume they are all technical—this would only set up the data science team for failure and the executive team for frustration. Here are some key challenges:
Mindset:
Companies have been saddled by unproductive thinking for as long as there have been companies. The many traps include data silo-ing; ‘not-invented-here’; ‘if it ain’t broke don’t fix it’; we can do it better manually; innovation doesn’t drive revenue; IT/DevOps/software/data science leadership can be covered by the CFO. None of these are true. You may have been able to get away with this before AI and data driven shifts, but this will not work much longer.
Hiring:
Getting the right people is tough, particularly when the market is tight for data science skill sets. However, it is only going to get (a lot) worse if you misunderstand what you need. Not every company needs the top AI/ML research team in the world. What is needed for most companies are well-informed, data savvy, business-minded leaders in all functions and teams that understand the value of the company’s data and bridge technical execution with the company strategy.
Rapidly changing landscape:
You will have to accept that it is impossible to keep up. This will create risk for your company. However, your failure is guaranteed if you stick your head in the sand and hope it will go away. Innovation must encompass planning and budgeting for tech. dev. research, assessment and on-boarding of new technologies, partnering with external expertise, and valuing adaptive processes above ‘hard-coded’ capabilities
Data Capture:
Your organization already generates enormous amounts of data by the standards of any prior decade. This growth is not going to stop, and if anything will continue on an exponential trajectory. Capturing and organizing data must become one of the company’s top priorities or the corporate machine will come to a grinding to a halt. The days of 20-year-old corporate software that was outgrown 19 years ago are over. If you are the decision-maker and you think this ‘ain’t broke’, then (unfortunately) you may be the most broken part of the system.
Digital transformation, especially now with the twist of AI, is daunting. This is certainly why many companies have decided (most unintentionally) to put the process off for years and in some cases decades. A place to start is to engage (and perhaps expand) the whole executive team. AI transformation requires center-of-excellence style thinking, incorporating leadership across all functions. In this new business environment, the risk is far greater if you don’t change