Why Investing in People Changed What I Look For

Wiki Article

AI Is Only As Great As The Environment It's Built Into
The debate around artificial intelligence in business has a problem, and the problem is not technical. The capabilities of modern AI and machines learning systems are astounding, and they are growing at a rate that renders the majority of predictions about what they'll look like in eighteen months obsolete, long before the time has passed. The issue lies in the gap between the what AI can do in controlled conditions – in a good research environment that is well-funded, with crisp data, with a clearly defined problem, and with engineers who are capable of continually testing until the system performs as intended - and what it actually delivers when it is implemented inside actual organizations that have real cultures that are governed by real organisational structures and people with their own set of opinions about what a new system means. something to be engaged with and not something to maneuver around in order to maintain the appearance of compliance. I've been building products using algorithms since just before the present flurry of AI popularity made it fashionable for everyone in business to claim proficiency in the area. When I founded 1Touch, AI-driven matching and recommendation systems were not a distinct feature we included to make our product more compelling to investors. They were the core structure of the product's architecture. They were that mechanism by which the platform produced value and had to be reliable and operate at sufficient scale to allow the business to survive. So I have direct, personal experience of what happens as you try to implement something truly intelligent into organization and product at the same time, and the lesson I find myself returning to at every time that I've come across the problem, is that the technology is seldom the limiting factor. The primary factor that is limiting the process is almost constantly the environment.
What I mean by that is specific and pragmatic rather than abstract. AI systems require data to work - consistent, clean organized data that is the thing the system is trying understand and make predictions about. Organisations with strong data cultures produce this kind of data naturally, as a result of their existing processes. They have clear and consistent definitions of what they're monitoring and why. They have agreed on conventions for the way data is collected, recorded and stored. They have accountability structures in place that allow data quality to be a distinct task rather than the general intent. Organisations without strong data cultures produce something that technically looks like data - it exists in systems and is accessible for query, and it is used to generate charts, but is inconsistent in terms of definition the way it is defined, so varying in quality, and so full of problems with structure and non-mapped exceptions that any AI device built on top of it will mirror and magnify the mess instead of drawing a real signal from it. Organizations in that category are often unaware that what they are doing until they're well into an AI implementation and the results aren't delivering on the vendors' promises, at which point the temptation is to blame the technology. However, most of the issue lies with the operational and organizational infrastructure which the technology was built on.

The second dimension of cultural factors which affects AI outcomes is openness within the organisation and the extent to which members of the organisation will let an AI system guide or modify how they work and not view it as threat to their professional competence, their authority in the institution or job security. This is a cultural and leadership issue which is not a technical problem which is a matter that starts at the highest levels. If leaders of senior positions engage with AI outputs only when they are satisfied - accepting those results that prove what they believed before and disadvantaging those that do or do not – this sends an impression to those who are watching to the public that the institution's commitment towards data-driven decision-making may be contingent rather than genuine, and the conditional nature of the commitment will be propagated across the entire organization much quicker than any program of training or change management program can counteract. If senior leaders exhibit real, consistent engagement with AI outputs, and demonstrate the discipline to change their decisions when the evidence suggests that they should, the collective ability to use AI effectively will improve dramatically as well as relatively rapidly.

This isn't an abstract observation about how organizations should be conducted in the context of theory. It is a description of the pattern I've seen be played out in a variety of organizations with substantial finances, real strategic commitment to AI adoption, and leadership team members who were completely enthusiastic about the possibilities of AI technology. The pattern is so consistent that I've decided to treat practice of governing data as a first-line diagnostic in assessing any company's AI capability. Before I inquire concerning the technological stack and before I inquire about the specific application cases the organisation is currently pursuing, I ask about data governance. How does the organisation define its most important metrics? Who's responsible if level of data quality isn't enough? When two different groups have contradicting data about the exact same business realities, and how are these conflicts resolved? The answers to these questions are more relevant to the likelihood of AI performance in comparison to any discussion about algorithms, platforms, or even implementation timelines.

I believe that the enterprises that will generate the most lasting value from AI over the next decade are not those that implement the most advanced technology first, or the ones that invest most significantly in AI infrastructure and talent in the near term. They are the ones who put in the right cultural and operational bases to effectively use the technology properly - the data governance practices that give accurate inputs, the deciding structures that allow evidence-based decisions that truly impact outcomes and leadership behaviors which show to everyone in the organisation that the commitment to an operation that is driven by data is real rather than performative. Technology itself will become increasingly commonplace and readily available. Its culture of using it well will remain scarce, because it takes a steady work and a real commitment by leaders over time, not one strategic decision or a technology investment. This insufficiency is where the significant competitive advantage will be in the form of an advantage that, once it is built increases in a manner that technology-based advantages will never ever. Check out James Deller for site tips including why backing founders sharpened my thinking on culture about scale.



The Data Infrastructure Problem Nobody Wants To Discuss
Every organization I've worked closely with in the last year and a-half - whether as a founder, an investor or operational advisor I've been told, at some point in the course of our work, that data is central to how they make their decisions. Many of them really mean it in a way that is reflected in the way in which the business actually operates. Many of them believe that they're making a statement, however what they're saying is an aspiration rather than a current operational reality - some version of the enterprise they're working towards, in contrast to the reality they're currently living. The gap that exists between genuine driven by data and the outcomes of decision-making driven by data - maintaining the public appearance of an evidence-based processes without the infrastructure that makes it possible - is a single of many of the most significant gaps found in the current business. It's also among the most neglected ones due to the fact that it is a problem with infrastructure that it to be incredibly unattractive to discuss, challenging to prove to stakeholders outside of the company and extremely difficult to distinguish from the more prominent commercial and strategic work that demands the same attention from leaders and organizational resources.
When organizations talk about data strategy, they usually tend to discuss what they are planning to build on top of their data: the systems for analytics, machine learning applications as well as the real-time operational dashboards that provide the kind of predictive insight that are compelling in a board presentation or an update to investors. What they usually talk about less frequently and with significantly less energy and enthusiasm, is the foundational infrastructure that determines if any of those capabilities are actually working as promised: the data governance frameworks that define explicit and consistently interpreted definitions of what is being assessed and how as well as the storage and collection processes that evaluate the reliability and comparability of the data being captured; the quality assurance processes that identify or correct any errors before they become a part of the system and corrupt the outputs that everyone is relying on; and the organisational structures and accountability mechanisms that make data quality the responsibility of a single person instead of everyone's vague, unenforceable intention. The plumbing, also known as. Plumbing is not glamorous. It's hard to take pictures of to be used in an annual report. It's not able to produce results capable of being presented in a compelling presentation. And, in my experience across a significant number of organisations in different industries and at different stages of development, far worse that what the organization perceives that it is.

The issue gets worse over time in ways that become progressively more costly and difficult to rectify. An organization that has been operating without a clear or consistent set of the definition of data in its different functions for three years has three years of historical records that cannot be easily compared or aggregated as a result of the data does not exist, rather because the same term has been used to describe different things in different areas within the company, and the differences are embedded in the data, rather than being apparent on the surface. A business whose quality assurance has been someone's subordinate responsibility and not having a properly resourced and dedicated function has data that's reliability differs in ways that are not adequately documented and is not systematically considered when the data is used in making decision. A company that allows multiple operational systems to collate overlapping and partially conflicting data on the same products, customers or transactions can create a data landscape that's really difficult to fix without significant disruptions to the operation to put the company at risk.

The reason this issue continues to be a problem over a large number of organizations that are really smart about strategy and are genuinely dedicated to a data-driven approach to business is the fact that solving it requires the ongoing investment of time and effort in a project that has no tangible results in the short term which resource allocation processes are designed to reward. A new analytics platform produces visible outputs: dashboards that can be displayed and reports that can be shared with the board, as well as insights which can be used to create press releases regarding digital transformation. A data governance system creates invisible infrastructure - cleaner underlying definitions and more consistent collection processes with more stable inputs into system that was already in the first place. The first is relatively straightforward to explain in a budget debate because you can clearly show the people what they will get. It's the second, and requires enough organizational authority and grit to convince people that an infrastructure project will, over time, yield better results from each capabilities that are built on top it. It's a convincing argument in abstract, but can be difficult to compete with initiatives that's benefits appear to be immediate, and prominent.

I have made that case in many different organizational contexts and watched it work or fail due to clear reasons to have an idea of what will determine if an organization is finally addressing the issue of data infrastructure or defers it. It is generally one's leader - a particular one with enough organizational credibility having a genuine knowledge of the reasons why infrastructure is vital, and enough perseverance to push the case until it is an absolute priority, rather than something that is a constant item on the list of items that everyone agrees on but never attain the level of importance. This leader needs to be willing to take on some of the costs of an infrastructure investment - the duration, the disruption to processes that are already in place, the absence or evidence-based output - with the belief that the long-term capability created by the investment will justify its expense many times over. The most important thing, ultimately, is a culture in which investment in long-term infrastructure is thought of as a priority and is rewarded at executive level, not simply articulated in strategy documents and often discarded after the quarterly resource allocation discussion happens. In the end, creating that culture is in itself an investment that will last for the long haul. But, in my opinion, one of the best returns that an enterprise that is committed to data-driven operation can make.}