Written by: Richard Frykberg

Artificial Intelligence (AI) is changing everything. What you do, and how you’ll do it. Machines have been replacing brawn. Now AI projects are replicating brain. Technology has progressed from being an obedient tool, to being an essential assistant, navigator and co-pilot. As the capabilities of AI rapidly evolve, so does the demand for AI projects to transform critical business processes. With limited expertise, infrastructure, time and funding, how do you effectively collate all this AI project demand, evaluate options fairly, and prioritize and deploy your most valuable AI projects successfully?

AI Project Transformation Opportunities

What can AI projects deliver? As it turns out, quite a lot.

Machines will never really be intelligent from a human, sentient, perspective. However, they have been trained on a LOT of data, and they have learned a lot. As the training data availability, techniques, and processing power have geometrically increased, the effectiveness of deep learning to derive the appropriate weights to apply to multi-dimensional inputs to produce meaningful outputs has dramatically improved.

The consumer-accessible evidence of these Large Language Models was brought to the fore with the public release of ChatGPT by OpenAI. This is now embedded in Microsoft Co-Pilot and broadly accessible via Edge and is embedded in Microsoft’s operating system and products. Similar Generative AI solutions include Google Gemini, Llama by Meta, and Claude by Anthropic.

The goals of all these models are description, prediction and prescription. What is it? What’s going to happen next? What to do about it?

Provided the input can be digitized, this type of feedback can be invaluable to the success of AI projects.

Example AI Use-Cases

In our daily lives we probably interact first with an AI system before any human. Every time we unlock our phones just by having them recognize us, we’re benefiting from AI computer vision and classification. Whilst access cards and PIN codes can be lost or shared, our biometrics are uniquely ours. Not surprisingly, therefore, use of AI capability for identification and security purposes is now highly prevalent. Whether by our face, our fingerprints, or voice we’re increasingly being seamlessly authenticated in all walks of life. Automated human identification is transforming point of sale, time and attendance tracking, and access control.

As humans our natural method of communication is not typing. It’s talking. Massive advances in Natural Language Processing (NLP) including speech recognition and speech synthesis open-up entirely new ways of interacting with technology. What would it mean if your digital sales agents could call out and capture orders without the need for your customers to access your website? How much more productive could field and maintenance staff be if they could just describe what they’re doing as they do it? How much time could be saved in consultations if the need to recall and transcribe conversations was eliminated?

Much human interaction with computers has been to obtain an analysis of past events. We’ve relied on IT to keep track of our expenditure, sales performance, connections and fitness statistics. Now, thanks to this large volume of accumulated history, and the application of machine leaning neural networks, the focus has shifted from descriptive analytics to predictive analytics.

Rather than relying on ever-optimistic human forecasts, AI systems have seen these patterns before and can provide a much less biased and more realistic prediction of future outcomes. What is the most likely forecast of your capital expenditure? Which of your product lines is most likely to be a hit this Christmas? Which customers are most likely to churn? And how could this predictive analysis help you avoid waste and maximize returns?

As AI models mature, they can handle ever more complex tasks. Already AI-powered software applications are automating routine tasks such as invoice processing, bank reconciliation, and material requirements planning. With chain-of-thought models enabling deeper reasoning and more accurate responses, AI is promising to handle much more sophisticated tasks. AI agents will start to manage customer profiling and collections, supplier risk assessments, staff selection, and product design. It is unlikely that current human involvement will be eliminated, as we will still be required to supervise critical actions, but our teams are likely to become much leaner and more productive.

Even in the physical world, we’re seeing a transition from manual to robotic operation. Coupled with computer vision, and reinforcement learning, AI-powered equipment is becoming rapidly more capable and cost-effective. Self-driving cars still crash, but it’s only a question of time before they’re considerably safer than human drivers.

Similarly, robotic manufacturing is getting better. AI never gets bored, sick or sloppy. Introducing robotic automation into manufacturing processes will inevitably improve quality, safety, productivity and agility. Manufacturing skills will need to transition from doing to training and supervising the machines.

Knowledge graphs help connect and relate organizational data with real-world facts to infer and produce valuable insights beyond the capacity of human recollection. These new insights can help uncover new product and market threats and opportunities.

As an entire organization transitions to an AI-powered future, the amount of digital data will only increase. Through the feedback loop of reinforcement learning, early adopters of AI will benefit from a virtuous cycle of continuous value realization.

What AI Projects Cannot Do (Today)

Unfortunately, or perhaps fortunately for us humans, there are some things that AI projects don’t do well.

AI scores very poorly when it comes to emotional intelligence. It is trying. For example, Microsoft (Oct 2024) has just released an update to co-pilot to be more friendly. “Over time it’ll adapt to your mannerisms and develop capabilities built around your preferences and needs.” Digital companions can help alleviate the scourge of loneliness in ageing populations.

However, no machine will ever provide the full human experience of intuition, empathy, joy and passion that define us. So, for the foreseeable future, personal services remain beyond the realm of AI.

Leading edge creativity and invention will rely on heavily on human participation at the boundary of our human-experience and exploration. However, AI is beginning to challenge humans in the arena of innovation: applying established technologies into new applications. Compiling existing images or music into new mashups and arrangements by employing GANs (Generative Adversarial Networks). Discovering and validating new combinations of materials and molecules.

Many organizations are seeing key opportunities for AI projects not in the replacement of current human involvement, but in the extension of human capability and capacity. To be able to grow without increasing headcount.

Urgent Adoption Imperatives for AI Projects

AI technology is unlikely to replace human ingenuity any time soon. However, its fair to assume, that AI-assisted humans will quickly replace luddites. We are entering an era where the transformational productivity benefits of AI-powered solutions will present an existential threat to many organizations that fail to respond to the opportunities presented.

A critical dimension to your evaluation of alternative AI projects is to rapidly assess the impact of AI on your product and services value proposition. How will AI impact your target market? Will it enhance or diminish your value proposition from your customer’s perspective? What are your major competitors doing with this new technological capability? What new entrants will this new technology usher in?

Knowledge-based services are most obviously at-risk. Call-centres are bracing for major upheaval. AI agents with access to all product documentation, previous call histories (including voice conversations), and train-of-thought reasoning capabilities will likely outperform their human incumbents imminently. They will be more responsive to the language and personality of their customers. Customers will increasingly prefer to be served by a virtual assistant than a harried and inexperienced outsourced human. And of course, they’ll be dramatically cheaper.

Advanced knowledge-based service industries are not immune. Medical, legal, financial, creative and IT services will all be heavily impacted.  These industries set a high bar for accuracy, reliability and liability. But humans are not invincible either. As the relative accuracy of AI-powered services improves, at dramatically lower cost, advanced services will be focussed on a vanishingly small share of premium service offerings for elite clientele.  After all, many of us would accept a 99% accurate response for a fraction of the cost, knowing that even the ‘experts’ don’t always get it right.

What about consumer products? We’ll always need food, clothing, housing and transport. What is the imperative for AI projects in traditional industries?

AI is not just a digital assistant for knowledge workers. AI projects are increasingly relevant in the real world too. Consider the rapid advances in computer vision and robotics. Machines are greatly increasing the productivity of agricultural producers. From genetic engineering, to weed and pest protection, to harvesting, storage, processing and distribution, AI capabilities are helping primary producers deliver more for less. Less energy, less harm to the environment, less waste, less cost.

Throughout all supply chains, AI projects are helping to more effectively customize offerings, predict demand, and optimize networks. Complex order management, production scheduling and fulfilment logistics are increasingly be streamlined through real-world learning encompassing a myriad of factors, rather than being reliant on the rules-based, data-constrained, Enterprise Resource Planning systems of the past.

So, the first question to answer for any variety of AI projects, is what is the risk of inaction? Where AI is already impacting your customer expectations, your service offerings, or already being rapidly adopting by current or new competitors, this urgency can be existential.

Business Benefits of AI Automation

In the current hype-cycle of AI enthrallment and panic, every manager of every department will be actively considering the opportunities and threats presented by this new technology.

The demand for adoption of AI capabilities will become insatiable. To ensure that limited IT, business and data science resources are allocated to the most valuable initiatives requires a fair and transparent evaluation of the anticipated business benefits of each initiative.

Business benefits ultimately come down to cost saving, revenue growth, and risk mitigation.

Cost savings can be achieved by increases in productivity that allow the same work to be done with fewer people. Inevitably there will be some job losses due to replacement of simple clerical tasks by AI agents. The labor-cost savings will start at the ground floor and work their way up through all knowledge worker roles (including IT, marketing, and design) and into the management layer which can commensurately be rationalized. Shop floor workers will be progressively replaced by faster, safer, and more efficient machines.

Time savings do not always translate to direct labor cost savings. Many organizations plan to redirect time savings to higher value tasks, rather than reduce headcount. These tasks may achieve even greater cost savings than the value of the saved hours.

For example, if the focus of the accounts payable team shifts from reactive responses to enquiries to more strategic vendor engagement, the benefits could come from better pricing, more reliable, in-full and on-time deliveries, and improved quality reducing direct costs and eliminating waste.

Adopting AI technology in marketing, sales and product design is likely to improve and protect revenue streams. Customer engagement, customization and service delivery can all be optimized through application of AI technologies to perceive and predict market demand.

Based on these insights, AI can help to refine an organization’s offerings. By accelerating communication flows, and expediting product and service delivery through automation, successful organizations will be able to more rapidly bring new and more desirable products to market.

A key value potential of adoption of AI is in risk mitigation. Fraud detection, quality inspection, and compliance are obvious examples where AI techniques can greatly expand the scope and frequency of surveillance. With access to broad data sets AI can do a more thorough job of vetting key business decisions from staff selection to CapEx approvals. AI in capital planning is able to cross-reference submissions to identify anomalies and provide a fresh, objective perspective to mitigate inherent business bias and data integrity risks.

Assessing Technical Feasibility of AI Projects

Alongside the potential benefits of AI-powered solutions, one needs to soberly assess its technical feasibility and implementation risk. The key constraints are performance accuracy, data accessibility, engineering complexity, and information security and governance.

The required standard of performance accuracy should be clearly specified by the request initiators. In regulated and critical business processes, for example in medical, transportation or agricultural industries where human life is at stake, a 99.999% performance accuracy may be required. If the available technologies in the target domain cannot yet reliably meet these performance expectations at commercially viable cost, these initiatives should be deferred.

Data is the lifeblood of AI. More data is always better. However, lack of sufficient training data should not necessarily be seen as an insurmountable obstacle to embarking on an AI transformation. The active AI ecosystem is rapidly producing both broad and highly specialized foundation models. Leveraging the power of these models together with an organization’s limited experience can still provide useful automation and insight.

The speed and effectiveness with which organization can consolidate their AI platforms, AI engineering capacity and data sources will help determine the technical feasibility of competing initiatives. For example, if you’ve already transitioned your Human Resource systems to a cloud service provider that offers AI capabilities, you may be able to achieve AI benefits in HR-related business processes easier than in other areas where you may only have traditional siloed data stores.

Where the proposed AI project relies on 3^rd-party processing or models, the confidentiality and security of any data exchange should be carefully considered. For routine tasks such as image to text processing, where the subject matter is mundane, this may be assessed as an acceptable risk. However, once sensitive personal or technical data is included, these information security concerns may present too great a risk to the organization.

Develop and Align AI Projects to an Informed Strategy

As with all investment initiatives, AI projects should align with an organization’s strategic goals and objectives. Good projects for one organization may not suit the current strategic imperatives of another. Given constraints, it is essential to clearly align AI projects with strategy to ensure that the projects that matter most are done first.

This may become an iterative loop: as an organization becomes increasingly adept at deploying AI capabilities, it may incorporate ever-more ambitious AI-related aspirations. As AI success begets business success, so will business success beget further AI investment.

The critical resource for AI success is high quality data. Once the AI flywheel starts to spin, and data is collected to deliver better products, that attract more users, that provide better data, it starts to spin faster and faster. This virtuous cycle of AI promises to deliver spectacular results for the pioneers.

Whilst only a few organizations will be born into the AI generation, many established industries will face an existential challenge: harness this new AI capability operationally and strategically, or run the risk of being left behind.

AI Transformation and the Race for AI Projects

AI transformation is vitally important, and you will have to start, even before you are quite ready. You will need to race to stake your claim over the critical data that will fuel your AI-engine.

Your choice of which direction to run may decide your victory or fate. AI investments should be treated like any other business critical initiative. The scoring model is less familiar, but an effective framework should be applied to make sure you pursue the initiatives that will matter most.

The key dimensions to evaluate for each AI project are its strategic alignment to organizational goals and objectives, its urgency considering competitive threats and opportunities, its relative business benefits, and its technical feasibility and implementation risks.

AI transformation has only just begun. Empower your team to embrace this new phenomenon, provide the tools to capture their insights, evaluate and rank these ideas consistently, and deliver your AI projects successfully knowing that your resources are being effectively committed to all the right areas.