Skip to content

NVIDIA, a maker of AI chips was worth around USD 300 billion in November 2022. Sixteen months later its value has soared by almost a factor of 8 to reach USD 2.3 trillion, making it the third-most valuable company in the world behind Apple and Microsoft.

One of the reasons behind NVIDIA’s astonishing growth in value is NeMo.

NeMo is an end-to-end platform for developing custom generative AI enabling enterprises to augment their large language models (LLMs) with proprietary data, allowing them to update a model’s knowledge base without the need for additional training or starting from scratch

This functionality in the NeMo service allows LLMs to retrieve information from proprietary data sources and generate human-like answers to user queries.

By comparison, an LLM such as Chat GPT, is trained on a diverse range of publicly accessible sources of information. Chat GPT does not access proprietary data and it does not retrieve data in real-time.

Last week at Bullhorn Engage the opening keynote from Bullhorn’s Jeff Neumann assisted by Nathan Green, gave me the light bulb moment that the NVIDIA investors had many months ago – productivity growth will be underpinned by LLMs that use proprietary data.

However, the extent to which this potential productivity growth turns into reality will overwhelmingly depend on the integrity and value of propriety data and the effectiveness with which the insights generated by the data are acted upon.

As Jeff and Nathan helped me see, the future for LLMs in recruitment can be summarised as Write, Find and Focus.

Write: LLMs can/will, for example, summarise the contents of a phone call and update the appropriate candidate, client or job record with this summary, write a job ad based on a job description and the accompanying recruitment brief; summarise a resume and interview notes into a concise and relevant candidate summary for a specific vacancy, craft marketing messages specific to defined customer segments, and compile a confirmation email from a phone conversation.

Find: LLMs can/will, for example, source candidates from the database relevant to the jobs added to the database; build a search list; bring jobs to the attention of candidates on your database and find relevant jobs your clients are posting (no matter where they are posted).

Activities under both Write and Find have the potential to save a lot of time however, the quantum leap in productivity will come in the third category of activities – Focus.

The activities I classify as Focus are ones where the allocation of resources, priority of action and elevation of skill have the potential to make a transformational difference in productivity.

Consider the following hypothetical examples;

a) A permanent job is added to the database then the LLM informs, based on comparable jobs, the chances of filling the job within 14 days are 25% however if the remuneration is increased by $15k the likelihood rises to 82%

b) A contract job is added to the database and the LLM informs you the client has an invoice payment average time of 47 days

c) A permanent job with a new client is added to your database and the LLM advises, based on Google and Glassdoor reviews and analysing candidate records, the new client has a 50% higher average staff turnover rate than the industry average

d) A candidate is added to the database and the LLM advises there is an insufficient sample size of comparable candidate records to predict the likelihood of placing the candidate

Based on the hypothetical scenario (a) the recruiter knows increasing the remuneration (or broadening the selection criteria) is a priority to increase both the likelihood and speed of a placement.

Based on the hypothetical scenario (b) the recruiter knows a conversation with the client about payment terms should be the next step in the process

Based on the hypothetical scenario (c) the recruiter will reconsider whether to work on the vacancy at all, and if they do work on the vacancy for how many hours and under what terms.

Based on the hypothetical scenario (d) the recruiter will know it’s a low-probability candidate and to manage the candidate’s expectations accordingly.

Competent and tenured recruiters will have a reasonably well-tuned radar for each of those hypothetical examples and the LLM will, largely, confirm their existing knowledge or instinct for what to do next.

It’s the rookie and low-tenured recruiters that have the most to gain from LLMs providing the above hypothetical insights. It will substantially reduce the necessity to learn through trial-and-error leading to higher productivity sooner in their tenure.

The other big winners will be team leaders who will spend less time coaching a recruiter on what to do and spend much more time on coaching them how to do the things the LLMs insights suggest are priorities.

The power of LLMs substantially raises the stakes in agency specialisation – the deeper the recruitment agency’s niche the more relevant and populated their proprietary data will be, the more powerful the insights generated by the LLM will be and the greater the potential productivity gain will be. And this advantage will quickly compound leaving competitors scratching their heads and wondering what happened to their market share – all within a very short period.

But it all depends on one critical thing – the integrity of the agency’s proprietary data.

If your culture is one of indifference to your own data’s integrity then any LLM insights are highly questionable because the data on which the LLM is drawing from is not codified, incomplete, and not current.

In the new age of AI and LLMs the future winners in the recruitment industry will be those agencies who combine a culture of codified, clean, and relevant proprietary data, with effective and targeted (by the LLM insights) skill development of their recruiters.

Related blogs

The AI end game for workers: greater focus on critical work

AI powers its way into recruiters’ jobs with GPT-4 (and its cousins)

A.I. will change agency recruitment but in an unexpected way

0 0 votes
Article Rating
Notify of

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Newest Most Voted
Inline Feedbacks
View all comments
Stephen Noble

Hi Ross,
Very interesting article thanks. I’m wondering whether Jeff & Nathan suggested that Bullhorn (and other CRM vendors) will be embedding NeMo into their CRM’s for their clients. Or whether agencies will need to employ the services of data scientists and implement NeMo themselves?


Stephen Noble
Australia Wide Personnel Group

Would love your thoughts, please comment.x
Scroll To Top