Getting DEEP with AI

By Gareth Parker, 

Chief Technology Officer, N3 Technologies (A Cornerstone Group company)

Generative AI is on the same curve as many previous groundbreaking innovations, and PropTech is adopting it the way every industry adopts a miracle: fast, wide-spread, and without pricing what comes later. I developed the N3 platform and run technology for Cornerstone Group whose primary objective is to be tech-enabled and trendsetting, so I am not arguing against it, however I am naming the problem that many are avoiding addressing. A catchy acronym and a clear problem definition can be managed so I call this the DEEP AI problem, and companies that have a strategy to address each of these problem domains will be the ones to succeed long-term on their AI adoption journey. 

Data. The moment a cost plan or a monitoring report passes through a frontier model, you lose control of two things: where the data went, and where that data is going. Many UK employees have used unapproved AI tools at work, mostly on personal accounts whose terms allow them freedom with your inputs, breaching GDPR and the Data Protection Act without the individual being aware of doing so. Tracking down these breeches is difficult and the AI that damages your business may not even be “yours”. It could sit inside a consultant's report or a subcontractor's tender, undeclared, and harmless looking. These are the “microplastics” of your data estate. The mitigation is rooted in data competency and awareness: Question your AI enabled tools (are they just cheap wrappers around American hosted providers?), control and verify your data sovereignty, monitor data provenance as part of your workflows, using AI to interface with your data, not generate it. 

Environmental. Data centers use a large amount of energy and draw in an incredible amount of water. For companies striving to be sustainable, AI is a conundrum. With the pressures of not falling behind, we want to use AI to replace mundane tasks, but in doing so we feed a resource hungry beast that is anti-thetical to a sustainable future. As responsible leaders we should measure the return on these tools honestly and commit a fixed share of the verified savings to sustainability, using the gains to help offset the hidden cost. This is something I would love to see PropTech to be the first to advocate and be an exemplar for the rest.

Ethics. Trust in our industry runs on the individuals who can be held accountable. Generative tools quietly removes the author from the picture. While it is already a legal obligation to verify AI output against authoritative sources; as the increase in AI generated material grows and becomes more believable, it becomes increasingly important that we disclose how and why we use AI in all our deliverables. Transparency is the only mitigation to this issue currently: say where AI was used, what was checked and by whom, and make that declaration as routine as declaring subcontracting or professional indemnity. While many tasks may no longer be our burden to bear, we do have the right to know what we are consuming and where it has come from.

People. Two anxieties are growing at once. People worry about their roles, and they are increasingly skeptical about the content they receive. Who wrote this, how, and can I rely on it? Our relationship with output is changing. The answer to the first is that the tasks AI absorbs, drafting, summarising, first passes, are the parts of most roles people would gladly hand over; judgement, accountability and knowing the client are the parts that make the role. The answer to the second is provenance: records that carry how they were produced, so that trust is something you can check rather than something you have to extend. Leaders need to build a framework to make sure we use AI to augment the human experience/expertise instead of replacing it.

Plastic took many years to become a crisis. AI is not a problem because it is bad, it is a problem because it is good. Addressing the DEEP AI problem now, while we are still exploring AI adoption, is the best strategy PropTech can adopt. Keep generative tools at the edge and deterministic systems at the core. Reveal where AI was used. Reconcile data against a source you can stand behind. Reinvest a share of what it saves.