How can AI help my consultancy?

I’m currently running a project with a few boutique consultancies and the Cardiff University Data Science Academy. As a result, I’m knee-deep exploring the implications of AI for consulting firms and professional services more generally. Joe O'mahoney Online Podcast

Given the space is pretty fast moving and there are new capabilities launched every day, I thought I’d put a line in the sand and put some key points.

The Current State of AI in Consultancies

We’re just getting started: GPT (using this as a short-hand for others!) are very powerful engines. However, the ‘car’ built around the engine that will get you places, is still pretty limited. There are a huge number of ‘parts’ being built every day, but these are very much in their infancy.

These parts make GPT more useful – for example, connecting it to the web or to your company’s documentation. However, the parts are generally been built by outside the GPT team and therefore are rarely as ‘trustworthy’ as GPT. The big firms are of course leading the way here, but have longer development cycles and are more risk adverse than smaller firms, so we will be waiting longer. For example, Microsoft’s Co-Pilot looks great, but I feel the best stuff will come from outside the big firms – time for some creative destruction!

Focus on the business: Like any tech, GPT is not a silver bullet. Any tech consultant will tell you that if you a hammer looking for nails isn’t an effective way to achieve business outcomes. For example, yes, GPT can save you a LOT of time, but then so can Excel short-cuts and .doc to .ppt exports. Think first ‘what do I want to achieve?’ and then think of the best tech (or non-tech) to achieve this.

There appear to be some common enquiries from consultancies coming into my inbox, which include:

Can GPT be connected to the internet ?

Can AI interrogate my firm’s documents?

Can ChatGPT bots answer common questions from clients and juniors?

Can AI interrogate my data better and faster than my consultants?

The answer to all of these is ‘yes, but‘. The BUT variously being related to limitations, ethics, security, or difficulty. I look at these in a little more detail below. But before I do so, I provide some more general observations.

The Future of AI in Consultancies: GPT4 and Beyond

GPT4 is a different ball-park: if you tried GPT3 when the hype was peaking and thought ‘no thanks’, it’s worth subscribing to GPT4 for a couple of months. Play with the Templates function and the other prompts [indicated by (k)]. The APIs are now much more robust and plentiful and can easily start to save your people time, especially those in marketing (see the blog and tweet functions in GPT4).

By the way, whilst we’re on the topic of marketing. One of the biggest challenges I see in outsourced (or even insourced) marketing is the production of high quality material that really speaks to client challenges. The issue here is that it is only usually seniors who can generate this type of content, but who has time!? I usually suggest that seniors write a summary (bullet points etc) and then juniors and marketing finish this off. However, now I would suggest that you experiment with GPT4’s ‘expand’ function to see if it can do this better!

The Impact of AI on Consultancies: Opportunities and Threats

Trust & Niche: The implications of AI for social media and digital marketing are remarkable. GPT4 can now generate hundreds of good blogs, tweets, articles, reports and presentations in a few hours (there’s currently a limit of 25 uses per 3 hours, though this will be increased soon). Most of this output will appear (though not may not BE) more insightful that those written by humans. Those teenagers who appear in your LinkedIn feed promising to scale your company to ‘8 figures in six months!!!!’? It’s all about to get a lot worse, and also more convincing.

Therefore, the old consulting adage about trust is about to get much more important. Trust in individuals will become crucial to a consultancy’s success as the gap between the “cheap but generic” and “personalised but expensive” markets expand.

This is also another reason to niche down in consulting: Whilst AI will soon provide cheaper (& sometimes more effective) advice on generic, low-risk, or ‘I need this fast’ problems, it will struggle to compete with tailored, important, and ‘I need the best answer’ questions. The more you can focus your solutions away from the generic and towards the specific, the less likely it will be that AI can compete with your insights.

Specific Applications of AI in Consultancies

Now let’s look at some more specific questions that I get from my consulting clients:

Can GPT use my .pdfs to learn?

Yes, you can feed GPT .pdfs from your company. You can use Langchain to convert and feed company pdfs to GPT and then run queries off them. These type of embeddings do not affect your tokens. However, it is not learning off your .pdfs, they are simply data on which GPT runs queries. You will need to wait some time before this is possible.

Can I use GPT to answer FAQs from my consultants / clients?

Yes. You can use your company’s relevant data (emails, chat logs etc) to create an index which is used as context information to provide responses. This is not a simple project though!

Can GPT search the web?

Yes and No. By itself, it cannot. However, GPTDiscord connected Google search and Wolfram this month! In short this means that GPT is no longer tethered to ‘what it was trained on in the past’ and can browse the web for more recent answers and also those which are more specific to your needs. The addition of Wolfram means that the analysis, presentation, and processing of the data that is found, suddenly became a lot more powerful.

The danger here for consultancies should be obvious (albeit in a few years) : if a client can ask AI to generate reports and answer questions and present the answers in beautiful charts, then 50% of consulting goes out the window. If personalised learning and coaching can be delivered by an avatar which has been trained on nearly all the data that is available, then that’s potentially another 20% gone…..

How should we progress?

Encourage your employees to experiment. Have a weekly / monthly get together to discuss ideas. You will need access to a developer (ideally python) for any company specific API/Embedding/Index work, however, there’s a lot that can be done.

As well as thinking ‘what can AI do to help our firm?’ also think ‘what can our firm do that AI will really struggle with?’. I fear that is the answer to the second rather than the first question that will really determine your future.

Visit link - https://podcasts.apple.com/gb/podcast/the-consulting-growth-podcast/id1607297792