Wadds Inc. works with ambitious creative and professional services agencies and communications teams dedicated to achieving growth with social impact.
Its focus is on helping management teams build a future proof, differentiated market position that exploits emerging opportunities, manages risk, embraces innovation and navigates economic volatility. Here founder Stephen Waddington looks into AI adoption.
We’ve called peak hype for generative artificial intelligence (AI) in professional domains such as marketing and public relations.
Everybody is talking about it, but there is limited evidence of anyone achieving anything like the 40% or more productivity gains promised by AI vendors.
This new class of technology is described as both eliminating professional work and capable of demonstrating emotion and intellectual reasoning. So far, the reality could not be more different.
The term AI itself doesn’t help. It’s artificial, and it isn’t very intelligent.
AI is based on training an algorithm to manipulate a data set called a large language model (LLM). It predicts words or phrases based on this data set.
AI has been ascribed a superhuman quality because it responds to natural language and can manipulate large amounts of data in real-time. Thanks to technology vendors such as Anthropic, Google, and OpenAI, AI is accessible to anyone with a web browser and internet connection at no cost or low cost.
It excels at pattern recognition but lacks comprehension of underlying concepts and cultural awareness. It supports reductive applications better than generative applications. Use cases in a professional domain include administration, content generation, research, relationship management and evaluation.
AI creates a series of societal and organisational risk issues, including copyright, data management, ethics and hallucination. Governance and training must be addressed before AI tools can be used commercially.
For all these reasons, organisations have yet to use AI systematically in a professional domain. Adoption is best characterised as experimental, although a turning point may be in sight.
A survey by Andreessen Horowitz, published in March, investigated the adoption of AI in large companies. It reported that organisations were moving from experimentation to rollout, including a broad application of generative AI and inhouse apps.
There are generally four approaches to the adoption of AI.
1. Limit the use of AI
This is a common approach in higher education and markets such as finance and healthcare, where the risk issues are deemed too great. A challenge is that individuals within an organisation may ignore governance and use it on devices outside the corporate firewall.
2. Workflow adaption
Professional workflow is deconstructed, and third-party AI tools are applied to augment tasks. Examples include transcription (Otter), image manipulation (Canva), research (Consensus) and reporting (Fireflies).
3. Democratise LLMs
LLMs are distributed to everyone within an organisation. The appeal of providing everyone in a team with access to an LLM is that it democratises access and allows users to find use cases within their own workflow. Microsoft is actively supporting this strategy with its customers with the launch of a tool called Copilot as part of the Microsoft 365 suite.
4. Custom LLMs
Organisations build their own in-house AI tool, typically by licensing an LLM API such as ChatGPT or using an open-source model to guarantee data management and security. An organisation’s IT team typically leads this activity.
AI technology holds substantial promise, but its practical application is a work in progress.
Mainstream adoption will require robust governance frameworks and training. Organisations must adopt a balanced approach, recognising both AI’s capabilities and limitations, to integrate this technology effectively into their operations.
If you run a creative or professional services agency and are looking for an experienced non-executive director to help build resilience and plan for growth, please visit www.wadds.co.uk