The AI Conversation Clouds Decision Making
AI is much discussed, but little understood. Often, people posing as authorities on AI are prolific social media users generating content with ChatGPT. Even respected institutions struggle to offer clarity when it comes to making real decisions about AI.
In this environment, it is vital that we develop robust and rational tools to aid our decision making. The AI typology and implementation framework offer one possible vision of the kind of tools we need to maximise the value in AI.
Maximising the Value of AI Means Looking Beyond Chatbots
Chatbots continue to dominate the AI conversation. Generative AI (GenAI), the form of AI upon which chatbots are based, has valuable applications across domains. This ranges from drafting emails to generating ad campaigns.
However, there is much more to AI than chatbots and GenAI. At an interpretive level, AI can be divided into four practical categories: deductive, sensory, kinetic, and synthetic. When assessing the range of AI types, it is useful to think in terms of this typology. I have summarized the meaning of these AI types in the following table:
| Type | Examples | Use Case |
| Deductive | Reactive AI, Predictive Analytics | Using (historical) data to produce insights for human decision makers. |
| Sensory | Computer Vision, Biometrics, Natural Language Processing (NLP) | Converting sensory data into computable data. |
| Kinetic | Robotics, IoT Actuators | Autonomously conducting physical operations. |
| Synthetic | Generative AI, Agentic AI | Creative interventions, or creating digital outputs. |
It is possible to combine different types of AI in what are known as multi-modal systems. As people and organisations are becoming better at utilizing AI, multi-modal approaches are becoming increasingly common.
The four-part AI typology, plus multi-modal approaches, creates a complex landscape that poses strategic difficulties for decision makers. We need a clear framework to enable us act in an ever more AI dependent global economy.
As in any high-stakes scenario, the decision to implement AI should be based on sound organisational, economic, and practical foundations. The need for this is perhaps even greater given the current state of the AI conversation.
AI Must Not Override What is Best for Your Organisation
Organisational and economic considerations apply to all decisions, not just those that pertain to AI. However, given the prevalence of “fear of missing out” and generally poor levels of understanding when it comes to AI, it is vital that we restate these considerations.
Your organisation must be able to afford any new project that it undertakes. This requires a qualitative assessment of your organisation’s capacity to reinvest in growth-focused projects. A composite of low debt levels relative to earnings, access to credit, and healthy cash flow are basic requirements for capital investment. It is also necessary that you conduct due diligence on external funding sources – circular financing has the potential to destroy your organisation and should be avoided.
Furthermore, any project that includes AI implementation must be subjected to economic scrutiny before it is underway. AI is not, ipso facto, value adding. You should require that any proposal to implement AI is based on a credible business case and a strong likelihood that your organisation, client, or customer will benefit. The temptation to implement AI simply because you can should be resisted in the absence of a sound economic basis.
Practicalities Can be Mundane but Are Vital When it Comes to AI
From a practical standpoint, AI delivers the most value where it works alongside humans. AI should be used to multiply the cognitive potential of humans, not as an act of beneficence, but because that is the key to unlocking maximal value.
First and foremost, AI has the potential to execute tasks that are practically impossible for humans, such as reviewing vast quantities of data or comparing data in numerous languages. AI implementation should start here. This is because the fastest route to (micro- and macro-) growth is to acquire previously impossible capabilities.
Secondly, as waste is a fundamental inefficiency, AI should never duplicate work done by humans. If there is a decision to be made about whether to use AI to execute tasks that can be done by humans, the decision should be to either replace human workers or retain human workers and accept the costs of doing so.
Choosing the latter option does not necessarily make you a luddite. There are often positive externalities to retaining human workers even when it is practically possible to implement AI. This is especially the case regarding tasks that are either partially or entirely based on interpersonal interaction.
We Need Robust Tools to Help Make Decisions about AI
Decision makers should use the AI typology and decision-making principles above as a framework for AI implementation. It is a tool that enables good decisions despite the noise in the AI conversation and complexity of the technology itself. Without such a tool, there is a significant risk of investing in low-return projects and leaving value on the table.
Implementing AI depends on a well-functioning organisation and should be based on strong projected benefits. It is vital these considerations are kept in mind, as they would be for any activity intended to develop your organisation. From a practical perspective, AI should either execute tasks instead of humans or, better still, expand your organisation’s capabilities into previously impossible frontiers.
