Melina Woollon, Head of Procurement at Bristol Waste Company and Director at Wallis and Woollon Consulting Ltd, explains why procurement process redesign in the waste industry is the key to escaping AI pilot purgatory.
According to recent research into Supply Chain AI readiness, there is a massive gap between individual AI adoption and the scaling of AI at the organisational level. Whilst most procurement professionals use AI at a generative level, less than 10% (the so-called ‘Performance Elite’) manage to push this to the next level, scaling up.
It’s not about better AI algorithms; it’s about designing intelligent value streams which can then be automated. In other words, before procurement professionals push the magic AI button, they should first wave the magic wand of lean thinking-centric process redesign.
This is particularly significant in the waste industry, where complex end-to-end supply chains complicate the AI picture. What’s simple, though, is the answer to why AI should be used: to drive efficiencies, increase productivity, and future-proof the sector by focusing on the highest value in the chain, whether this is in the daily collection of municipal waste or the material recycling and disposal aftermarkets.
Similarly to most industries where AI adoption has been relatively late and marginal, the use of AI by individuals may increase individual productivity but does little to nothing to increase the end-to-end company performance.
In the UK waste industry in particular, the lack of success scaling up AI can be explained to a degree by the heavy reliance on a traditional mix of machinery and labour in a sector where the residential waste collection remains the statutory duty of budget-restricted public sector organisations, such as councils and their Teckal arms.
Instead of waiting for AI to solve organisational inefficiencies, waste industry procurement teams should replicate the Performance Elite’s modus operandi: simply put a new operational architecture in place ready for automation.
We absolutely need to slow down first (with a metaphorical whiteboard and marker) before we accelerate – otherwise, we risk making inefficient processes exponentially quicker. Mapping how waste streams flow throughout the supply chain – for example, from the collection point all the way to after-markets is a key starting point for redesigning this for everyone involved.
Redesign first, automate after
It’s that simple, just as the classic lean thinking methodologies have always claimed. To avoid yet another AI Pilot purgatory and the chaotic chase for yet another automation exercise, waste management teams – whether operations-facing, such as the Collection teams or support functions such as finance, procurement or sustainability – should ensure that a robust AI scaling framework is in place first.
This absolutely needs a planning phase: the better the planning, the better the result. We also need to think hard about the pilot and refining stages and how they all flow into the operationalise and scale phases. Remembering to include key milestones, such as strategic alignment and stakeholder engagement, is fundamental, as well as being clear on how to measure success and how to formalise governance. Without this type of discipline, we will be forever stuck in another AI pilot purgatory.
In practice, we also need to learn from those who tried AI without a framework and accept that sufficient insight is now available: without this, the operational scaling simply failed.
According to research, among organisational functions, procurement and supply chain show the widest gap between the ambition of AI scaling and its execution: despite just over 10% of respondents having no plans, only a small 4% are operating at scale. Interestingly, the perception among executives is that procurement has been the most successful of AI use cases. We need to be honest about where we are and correct the narrative.
The good news is that some of the key waste management processes to which procurement is key, such as waste containers ordering, PPE demand planning or the daily waste and recycling collections, have higher scale capability by nature, which is most likely linked to greater predictability. Interestingly, analysts conclude that the hardest-to-scale cases are those which need a high degree of human interaction, which makes the ‘last mile’ challenge an operational, not a technical one.
For example, leveraging spend category insight and throwing spending patterns analysis into the mix can significantly add value to procurement spend analysis. But the highest value will be achieved from the transaction-heavy process redesign, such as inventory management or ordering.
This cannot be stressed enough: if there are no formal governance frameworks in place, most departments, and procurement is no exception, will fall into the trap of following top-down instructions and risk chasing half-baked ideas – the use of best practice insight into waste management or relevant research should be used to offset this. The key is combining bottom-up insight with a top-down governance framework for a powerful business model which works day in, day out.
Escaping the AI pilot purgatory
AI requires a different rigour in terms of process and different skills to achieve this. There is undeniably a significant gap between the rationale of operational processes (what type of waste needs to be collected and how often), waste category intelligence (understanding the wood collection and processing industry outlook before articulating a response to this), and the real comprehension of AI models (in terms of their strengths, but also their limitations).
Unfortunately, a lot of AI users struggle to understand the mathematical models behind AI, which is why they tend to use this in areas with a perceived lack of human expertise, such as in the ever-changing waste legal and regulatory landscape.
If AI can analyse huge amounts of data and articulate an almost instant response, this can also mean the goal of key workstreams can (and should) evolve rapidly; for example, not just forecasting waste collection demand, but influencing it by deciding how to redistribute waste vehicles and labour resources intelligently to achieve more with less.
A new lean mindset is therefore needed, but the principle of this may be familiar: if employees normally learn from a few limited examples and generate ideas worth scaling up, AI, on the other hand, needs a lot more examples to refine and shape its behaviour and, unlike humans, may be unable to justify its reasoning.
Some key strategies can be deployed to ensure AI is optimally scaled up in the waste industry. When looking at decision risks, a powerful strategy is to try and balance human and AI oversight. If we were to look at the human intervention cost and compare this to the cost of good or bad decisions as an outcome, there is a case to be made for focusing on what some experts call asymmetric risk.
This means that we need to ensure human intervention is scoped in at those pre-defined thresholds where the cost of being wrong, for example, is deemed too high. A clear example of this is the H&S non-compliance risk where the cost of human intervention is low, but the cost of a bad decision is huge and may cost lives.
Other successful strategies may also look at stakeholder management and talent management (we may call the latter talent density creation). As waste industry leaders and procurement experts, we must dedicate focused time to redesigning our processes, and not just expect AI to automate them, and we should bravely look to build new skills and capabilities internally.
Nowadays, anyone can create a good spreadsheet using AI; however, we still need humans with the soft skills required to run an effective waste disposal strategy meeting and get everyone on board with a new market solution.
Ultimately, the waste industry can benefit from AI not because it articulates more intelligent responses, but because it can suggest more intelligent actions within existing processes, which are not easy to refine – these actions can in turn be replicated at speed and scale.
This is no easy task: it requires the discipline to integrate AI and stabilise this further to ensure AI is only used once the right controls are in place and we can confidently understand and trust an AI model to move us from prediction to influence and ultimately change.
This can only be achieved if we think about improving our key waste value streams first rather than chasing up the latest, shiniest AI pilot to deploy in a silo and with little impact on the larger strategy.
The difference between any Procurement Performance Elite and the rest may well be in the former’s ability to simply not chase the next AI model, but to build better processes beforehand. And to ultimately run better waste operations which can in turn unleash more value from anyone in the chain. In other words, to do the thinking.