Covelent: The balancing act of GenAI adoption in consulting

Transformative promise

GenAI itself was not new. The research that laid the foundation for the development of the models behind GenAI started in the 1930s. The first chatbot, ELIZA, which could simulate basic conversations with a psychotherapist, was developed in 1966. However, recent developments in AI technology had made GenAI more powerful, more versatile and more accessible to ordinary people. Earlier forms of AI had been good at routine tasks, but the release of ChatGPT in November 2022 showed that AI was now capable of creative, analytical and writing tasks. Moreover, the technology required little skill on the part of the user and it could achieve output that was almost indistinguishable from that produced by humans.

GenAI’s uses were not confined to text generation and writing (Bard/Gemini, ChatGPT, Jasper and Perplexity AI, for example). It could also be used for image generation (Dall-E2 and StyleGAN), coding (AlphaCode and GitHub), video generation (Synesthesia, Haygen) and a mix of these (multimodal). Not only could GenAI models generate text, pictures, audio and video, they were also being used in the development of new medical drugs.

Developments had not slowed since the launch of ChatGPT 3. Instead, the technology was changing so rapidly that each milestone was bringing the world ‘closer to a future where AI seamlessly integrates into our daily lives, enhancing our productivity, creativity, and communication’ (Marr, 2023).

The promise for business

Research by global firms such as McKinsey, Goldman Sachs and Deloitte was showing that organisations expected GenAI to transform the way they did business in the near future (Dutt et al., 2024; Goldman, 2023; Lamarre et al., 2024). McKinsey observed that GenAI had ‘risen from a topic relegated to tech employees to a focus of company leaders’ (Lamarre et al., 2024).

Other commentators noted that GenAI had ‘great potential to contribute to different types of value creation mechanisms, including knowledge creation, task augmentation, and autonomous agency’ (Feuerriegel et al., 2024:119). These authors saw the use of GenAI leading to ‘the development of new business ideas, unseen product and service innovations, and ultimately to the emergence of completely new business models’. They believed that in the future, the nature of work was very likely to change at all levels of an organisation and they asserted that GenAI would have the most dramatic impact on domains and industries that relied on creativity, innovation and knowledge processing.

Such was the scale of the predicted impact of GenAI that Goldman Sachs, estimated that over the next 10 years, AI could increase global productivity by 1.5% per year, resulting in a 7% increase in global gross domestic product (Goldman, 2023).

Cautious adoption

International IT consulting firm Gartner had developed the concept of a hype cycle for emerging technologies that tracked attitudes towards new technologies over the early stages of their life cycle. The cycle started at a low point with an ‘innovation trigger’, and then proceeded up a steep curve to a ‘peak of inflated expectations’, then dipped to a ‘trough of disillusionment’, before increasing slightly to the ‘slope of enlightenment’ and finally the ‘plateau of productivity’. In late 2023, Gartner placed GenAI at the peak of inflated expectations. The company made the following observation: ‘Even with robust pilots, broad experimentation, enthusiasm, and planned investment, the pace of adoption is forecasted to slow, at least in the near term’ (Bant et al., 2023). Gartner nevertheless predicted that GenAI would start to have ‘transformational benefit’ on organisations, even as the technology entered the trough of disillusionment.

By the summer of 2024, Gartner’s predictions about adoption of GenAI seemed to be accurate. As trust in the technology and its implications in practice was still low, actual use of Gen AI within organisations was not yet widespread, and its uses tended to be limited to areas of low risk (Hayden, 2023). ‘If you were to look at those organisations that have successfully integrated GenAI, it’s quite a small number in the grand scheme of things’, observed Nicholas. ‘And if you were to look at the subset that have integrated it in a meaningful way, then there are even fewer. It’s actually quite limited to companies like Microsoft and Google that have got billions [of dollars] to invest into actually integrating GenAI and building products around it’. He believed that there were a number of variables that influenced an organisation’s enthusiasm for implementing GenAI: the size and age of the organisation, the vision of its leaders, the region in which it was based, and the sector in which it operated.

Uncertainty and concern

Data security was one of the biggest concerns that organisations expressed when it came to using GenAI. According to Nicholas, while the financial services sector was typically quite technologically advanced and therefore more ready to adopt new technology, many financial institutions had banned their employees from using generic GenAI to ensure data privacy and data security. Not only was it possible that personal information could be made accessible to GenAI applications, it was also possible that confidential company business could enter the public domain.

‘If, hypothetically, you're working for a big bank, and you leverage GenAI for valuation assistance to help you value a Coca-Cola group brand because you are working on a potential sale to PepsiCo, that query has now been ingested by OpenAI. Therefore, there is a potential that somebody else could get that information. And if this happened, it would be disastrous for the bank and all parties involved’, he explained. ‘So there are those companies that completely ban it until they can understand if there is a way for them to integrate GenAI into their business in a safe and secure manner’.

Ethical concerns

Aside from data privacy and security, all the ethical grey areas relating to AI in general, as articulated in the EU Ethics Guidelines for Trustworthy AI (EU, 2019), were relevant to GenAI:

• Its potential to perpetuate bias and stereotypes and to amplify stereotypes if the deep learning models on which they were trained were based on biased data.

Practical concerns

There were other concerns that related specifically to GenAI. Among these was the reliability of the content produced. Much had been spoken about its tendency to hallucinate – to produce incorrect results, and to make errors in maths and when providing citations (Feuerriegel et al., 2024). Respondents in McKinsey’s research considered inaccuracy to be the most relevant risk related to GenAI. Less than half were convinced that their organisations had measures in place that could mitigate this risk (Lamarre et al., 2024).

Dell’Acqua et al. (2023) observed: ‘Some unexpected tasks (like idea generation) are easy for AIs, while other tasks that seem to be easy for machines to do (like basic math) are challenges for some LLMs. This creates a “jagged frontier,” where tasks that appear to be of similar difficulty may either be performed better or worse by humans using AI’. Their research found that within this frontier, GenAI had the potential to complement or displace human work. Outside the frontier, GenAI output was less useful because it was inaccurate, and as a result it could degrade human performance.

‘Because the capabilities of AI are rapidly evolving and poorly understood’, they concluded, ‘it can be hard for professionals to grasp exactly what the boundary of this frontier might be at a given point in time. We find that professionals who skilfully navigate this frontier gain large productivity benefits when working with the AI, while AI can actually decrease performance when used for work outside of the frontier’ (Dell’Acqua et al., 2023).

Nicholas believed that there had been a degradation in the quality of the output of GenAI. ‘There are vast differences between the different platforms, which makes it very hard to choose to use another one if one model is not working. Some are far better at some activities than others. And the time saving that you would have obtained is degrading over time’, he said. Thus, while he found that using GenAI still saved him time, it was saving him less than it was a year ago.

In his view, another challenge was that ‘60% of the content that is out there now online is AI-generated. A significant amount of recent content is written by AI, and therefore, the models that are now training themselves on new data are just training themselves on their own outputs, which are flawed because they’re being written by an AI anyway’.

He continued: ‘When we're talking to clients, if we're taking secondary data from the market, this needs to be documented, so that we can build our assumptions and clearly delineate to a client the information on which we built our assumptions. But the process of getting the source out of these models is becoming harder and harder. And now it’s almost come to the point where one might as well just revert to using Google’.

People impact

A final area of concern related to the impact that GenAI would have on organisations from a people perspective. Goldman (2023) predicted that the jobs of hundreds of millions of knowledge workers would be replaced by GenAI. Respondents in McKinsey’s survey anticipated workforce cuts in certain areas and that large reskilling efforts would be needed to address shifting talent needs (Lamarrre et al., 2024). Others asserted that implementation of GenAI would have a strong impact on work patterns, organisational structures, leadership models, and management practices (Feuerriegel et al., 2024).

Covelent’s study

To get deeper understanding into the thinking of UK professionals, Covelent conducted a Workforce Survey that involved more than 2000 professionals from a variety of industries in the UK. Again, trust emerged as an issue. ‘The findings paint a picture of a professional community at a crossroads’, said Covelent’s report. ‘While there is recognition of the potential benefits AI can bring, such as enhanced efficiency and innovation, there is an escalating concern, with 65% of workers apprehensive about the rapid pace of AI development’ (Covelent, 2024).

One of their main fears was that their jobs might be automated more quickly than anticipated. Only a quarter of respondents felt prepared for AI integration. Those in the manufacturing and financial services sectors were particularly uneasy about the possibility that their jobs might be automated.

Ethical considerations also emerged as a concern. This was especially so in the healthcare sector, where professionals feared algorithmic bias in medical diagnoses, and in finance, where professionals noted the opaqueness of investment decisions that were based on AI. Professionals wanted to avoid potentially unpleasant repercussions if GenAI was deployed rapidly and without adequate ethical frameworks in place.

Respondents to the survey were also concerned about their own ability to adapt to a decision-making process that used GenAI. They feared a loss of personal judgement and expertise.

A final and unexpected finding was that respondents believed that GenAI might have a negative impact on diversity and inclusion in the workplace. Almost half of the respondents believed that if AI was not properly managed, it could perpetuate existing biases and even create new forms of inequality.

‘The survey highlights a crucial shift in the professional perspective towards AI. The accelerated pace of AI development has not only fostered optimism for innovation but has also intensified concerns about job security and the ethical dimensions of AI in the workplace. This evolving sentiment underscores the need for a more strategic approach in preparing the workforce for an AI-dominated future’, said Covelent’s report.

The findings in Covelent’s research were therefore consistent with the findings of the research conducted by many of the global consulting firms, which could be summed up by conclusion of the Deloitte report: ‘Talent, governance and risk are critical areas where GenAI preparedness is lacking. Will GenAI be the greatest, most impactful technology innovation in history? Will it completely transform how humans live and work? Or will it turn out to be just another technology du jour that promised revolutionary change but ultimately delivered only incremental improvement? Right now, we can’t be certain. What we do know is that many breakthrough technologies of the past have followed a common adoption pattern: initial awareness; excitement that led to hype; mild disappointment as hype met reality; and then explosive growth once the technology reached critical mass and proved its worth. GenAI seems to be following the same pattern, only much, much faster’ (Dutt et al., 2024).

Based on its own findings, Covelent had four recommendations for companies wanting to deal successfully with Gen AI:

1. Embrace upskilling and reskilling as core strategies and acknowledge that AI-related training is imperative. Such training should cover not only technical education, but also the ethical, societal and practical implications of AI in specific sectors.

GenAI opportunities within Covelent

When Nicholas considered how to use GenAI in his own business, he wanted to avoid getting taken in by the fearmongering associated with the new technology. ‘Fear sells. Every day we read articles and perspectives on LinkedIn talking about how GenAI is going to completely get rid of the professional services industry, particularly in consulting. And it’s always interesting reading. But I don’t think we should take this at face value’, he said. He gave the example of the response to the launch of Microsoft Excel in the mid-1980s. Then there were fears that it would bring about the end of the accounting profession. Instead, Excel had become an integral tool of the profession, augmenting and empowering the work of accountants and bookkeepers globally. He believed GenAI held the same promise.

He believed that a key to ensuring that Covelent maximised the positive impact of GenAI might be to think about it in terms of integrating technology, data, people, processes and ethics (see Exhibit 2).OPEN IN VIEWER

Nicholas was already deriving benefit from using GenAI. He no longer used Google as a search engine. Instead, he used ChatGPT or Gemini. He also used GenAI to assist in some of Covelent’s consulting work. He gave an example: ‘I have asked either ChatGPT or the Google Gemini version to come up with a valuation and then compare that to other valuation models. Typically, we would build our own Excel models and then compare that with others. But you can say to ChatGPT: “I would like to build this model. Can you let me know what inputs you need from me?” Then I can provide those inputs and ask ChatGPT to build a model and do the calculation for me. And it takes you through it step by step – for young aspiring consultants you don’t need to know how to build the model yourself. Of course, it helps to have a comprehension of what the model does, but if there’s a term in there that you don’t understand, you can ask for an explanation’.

In other instances, Covelent used GenAI to analyse and cross-reference client business cases with the business cases that existed in the public domain. ‘We’re asking whether the work that our clients are asking for is unique. We're also using it with our own data to feed in a potential case and find out if we have any existing data points from our own research that will underpin this particular case’.

He saw tremendous possibilities in GenAI for Covelent as a business, based on potential time saving alone. ‘It's turning tasks that may take a couple of hours into ones that takes half an hour. When you scale that up across teams, you could potentially be cutting down a 12-week project to 8 weeks. For a client, this has expedited their time-to-market, and the speed at which they can make informed, data-backed decisions. This also has commercial benefits for the client and value creating ones for the consultant’.

He also saw its potential for knowledge management within Covelent. One of the problems every consultancy faced was that of transferring the knowledge of experienced consultants to those who were newer to the profession. Conventionally, the less experienced consultant would have had to ask the experienced consultant for help or physically trawl through documentation to find what they needed. Now, it would be possible for a proprietary LLM to search for the relevant information.

‘As we grow as a business, the more research that we have, the more research we can feed into our data bank and the more information and insight we can get out at the end. It will likely never completely remove the dependency of having a senior person involved in that opportunity. But it does mean that we don't have to set up a meeting with somebody internally to ask them about that. So we might save half an hour actually physically querying someone, because we'll be able to get that information via the LLM that we've developed internally’, he said.

How to advise clients regarding GenAI?

When clients approached Covelent for advice, one or their biggest questions was whether they needed an AI strategy. Nicholas was not convinced that this was the most pertinent question. In his view, AI was an enabler of business strategy and should therefore not be the subject of a strategy of its own.

As a new consulting firm, Covelent had to be sure to add real value to its clients and Nicholas was aware that he was operating in an environment where the big consulting firms were putting huge resources into consulting on AI. He had to stand out and there was not much margin for error. ‘As a strategy house, we had almost zero brand reputation behind us as a collective’, explained Nicholas. ‘So for me, it’s about building up a bank of really well-delivered work that drives value’. He wondered how he could ensure that his firm delivered on this objective in this landscape of constant change, grand promises and low trust, and at the same time maximise the potential of GenAI in his firm.