#20 AI Wrecks Recruitment: The Unintended Damages
Plus: Buy or Build - AI Challenges SaaS, OpenAI's o1 Release, Extraordinary Valuations for AI Startups Founded by Extraordinary Scientists, Google NotebookLM Upgraded, and More
In this issue, I cover:
Notable News and Developments: OpenAI’s o1 release, highly valued AI startups, MIT’s AI risk repository
Buy or Build? AI Challenges SaaS
AI Wrecks Recruitment: The Unintended Damages Assistants Deliver High Returns
AI Tool Spotlight: Upgraded Google NotebookLM as a Learning Tool
One More Thing: Opt out of LinkedIn AI training data collection
Enjoy.
Notable News and Developments
1. OpenAI’s o1 Release: Think before You Speak
OpenAI launched its new AI model family, o1, on September 12, 2024, which includes o1-preview and o1-mini. These models are designed for advanced reasoning capabilities, enhancing problem-solving in areas like science, coding, and math. The o1-preview performs at a PhD level in some tasks and employs reinforcement learning with a "chain of thought" processing approach.
This launch is highly anticipated as it represents a major advancement in AI reasoning, potentially transforming complex problem-solving across various industries, though it comes with higher costs and slower response times.
So far I haven’t seen compelling business use cases and will keep you informed if they emerge. OpenAI also kicks off its newest fund raise valuing the startup at $150 bn according to some reports.
2. Extraordinary Valuation for Extraordinary Researchers
Recently we are seeing a few AI startups getting extraordinary valuation on their founders’ extraordinary reputation in AI research. Two examples:
Dr. Fei-Fei Li, the ‘godmother of AI’ and renowned Stanford professor, has raised $230m for WorldLabs, a startup with 20 people working on AI that can understand the 3D physical world.
Ilya Sutskever, OpenAI's former chief scientist , has raised $1 billion in cash for his startup SSI with 10 people, to help develop safe artificial intelligence systems that far surpass human capabilities. The three-month old startup is now valued at $5bn.
If we assume these founding teams give out 15% of equity to investors, and VC investors expect similar 100x return target for startups before product-market-fit, these companies will need to reach hundreds of billions in valuation to ‘meet the expectation’. Or, is this no longer the right framework to think about VC investing?
3. MIT’s Comprehensive "AI Risk Repository"
The repository provides:
An accessible overview of the AI risk landscape
A regularly updated source of information about new risks and research
A common frame of reference for researchers, developers, businesses, evaluators, auditors, policymakers, and regulators
A resource to help develop research, curricula, audits, and policy
An easy way to find relevant risks and research
It is interactive and searchable. If you are working on risk frameworks tailored to your industry or organization, this is a great resource to refer to. For board directors, this could be a resource for deep dives on any specific AI risks arise in your boardroom discussions. Here is a short video explaining how to use.
Buy or Build? AI Challenges SaaS
Klarna's recent announcement has stirred the SaaS industry. The Swedish fintech's CTO, Koen Köppen, stated: "We are sunsetting Salesforce and replacing it with an in-house solution. We're also in the process of replacing Workday." This move could save Klarna millions annually while offering more tailored solutions. It also hints at a potential new revenue stream – companies could potentially sell their custom-built solutions to others in similar situations.
The "buy vs. build" debate in enterprise software is not new. But it is shifting to a higher gear due to huge improvement in AI tooling in recent weeks. Coding tools like Cursor and Replit are turning everyone into software developers and raised productivity of in-house technical teams significantly.
Although very limited near-term impact, continuous improvements in the future could prompt companies rethink their IT budget and challenge established SaaS models. Customer churn rates could increase and revenue retention could become more difficult. The SaaS business model might be under threat in the long run.
The "build" approach has many challenges. Internal tools often falter during maintenance and upgrades due to insufficient product management. SaaS vendors have accumulated extensive workflow knowledge and can develop sophisticated products based on insights from their customer base. This specialized expertise is not easily replicated in-house. Moreover, developing custom solutions may divert resources from a company's core competencies, a concern for many investors who favor strategic focus.
Success in building custom solutions depends on certain factors. Companies with strong engineering teams and adaptive mindsets are better equipped to develop and maintain in-house software. This approach can significantly reduce costs, especially for large enterprises with substantial SaaS spending. The key is balancing the potential savings against the risks of moving away from tried-and-tested SaaS solutions.
However this trend may still reshape the SaaS industry. Providers might need to offer more flexible, customizable solutions to stay competitive. Companies could shift budgets from SaaS subscriptions to internal development and AI tools. SaaS companies may need to emphasize their unique value propositions and industry-specific expertise to differentiate themselves. The pricing model may change as a result. According to this article on CIO.com , Salesforce is mulling a new pricing model for AI agent conversations based on conversion success rather than seat or usage. This direct linkage to ROI is good news for its customers, but calls for fundamental changes in Salesforce’s own operational structure and unit economics.
Looking Ahead: The Future of SaaS SaaS providers may need to offer more flexible, customizable solutions. Companies could shift budgets from SaaS subscriptions to internal development and AI tools. The Hybrid Future Future enterprise software may blend SaaS and AI-assisted custom development to meet diverse business needs.
AI Wrecks Recruitment: The Unintended Damages
Is AI killing the job market?
The job market's looking a bit off-kilter these days. Indeed’s Job Posting Index shows we're back to pre-Covid levels, down about 30% from the 2021 peak.
You might think this means employers are having an easier time finding talent.
Think again.
The AI-Induced Hiring Paradox
For white-collar jobs, it's a different story. The openings in professional services and software developments have fallen on business priority shifts and reversal of ‘over-hiring’ in Covid. Rethinking of staffing needs in the age of AI contributes to the pullback.
As reported by the Financial Times, job hunters are twiddling their thumbs waiting for interviews, and when they finally land one, securing the job feels like pulling teeth. On the flip side, employers are drowning in a sea of candidates, struggling to fish out the right one.
AI Makes Resumes and Cover Letters Useless
AI has significantly altered the landscape of resume and cover letter evaluation. It has inadvertently compromised the effectiveness of these traditional tools in matching talent with roles. AI's capability to enhance every application, sometimes beyond an accurate representation of a candidate's experience or skillset, has led to a homogenization of resumes.
This widespread use of AI in application preparation has resulted in a flood of seemingly perfect candidates, making it increasingly difficult for employers to discern genuine qualifications and experiences. The traditional differentiators in resumes and cover letters have lost their efficacy, creating a challenging environment for both job seekers and recruiters.
The situation has reached a point where these once-crucial documents now provide limited value in the selection process. Employers find themselves grappling with an overwhelming number of ostensibly qualified candidates, while truly suitable applicants may be lost in the shuffle.
The Vicious Cycle
Recruiting managers, faced with this mountain of sameness, turn to AI for help. But here's where it gets messy. AI algorithms favor candidates that tick all the boxes - quite possibly the same candidates whose resumes were buffed up by AI in the first place. It's a vicious cycle. Candidates, getting radio silence on their applications, use AI to apply for even more jobs. Employers, swamped with applications, lose confidence in who they're actually interviewing and hiring.
Recruiting Struggles to Adapt to Changing Skill Requirements in Roles
Here's something I hear again and again from strategic leaders in the age of AI: they want to emphasize soft skills and mindset such as curiosity, fast learning, problem-solving, storytelling, and strategic thinking in candidates. The technical aspects of roles are either unnecessary anymore or easier to acquire with AI.
But here's the kicker - when faced with a sea of sameness, busy hiring managers fall back on pattern recognition. They end up hiring based on specific experiences that fit the box, which is the opposite of what they set out to do.
The recruitment model, as we know it, is broken.
What Comes Next?
Could this make traditional human-centered executive search more appealing, even for mid-career professionals? Maybe. But those relationships take time to build, and many recruiters trained in the ATS era might not be ready for this shift.
Talent recruitment in a full-blown identity crisis. AI's shaken things up, and we're still figuring out how to put the pieces back together. Until we do, both job seekers and employers are stuck in this weird limbo, trying to navigate a system that's not quite working for anyone.
Is the CV/resume era over? It's tough to say, but one thing's clear: we need a new approach to matching talent with opportunities. What are your thoughts?
AI Tool Spotlight:
Upgraded Google NotebookLM as a Learning Tool
Back in February, the third issue of this newsletter introduced Google NotebookLM as a writing and research tool. Google has recently rolled out a significant upgrade that has transformed this free tool into a powerhouse for both writing and learning. I have been recommending it to busy board directors to help accelerate their learnings on newest regulations and guidance, presentations, white papers, and industry research reports. [PSA: Do not upload your board meeting materials into any public AI tools. ]
NotebookLM now allows users to upload files, refer to internet sources, and create personal notes into the workspace. Among the tools, people love its ability to create a podcast-style audio overview of the materials in a remarkably natural way, particularly beneficial for auditory learners.
To demonstrate, I uploaded a 106-page white paper from the CFA Institute on private credits, a topic I’ve been studying. NotebookLM could create a summary and allowed me to ask questions, create my own learning guide and quizzes, cross-reference study notes and generate an informative audio overview.
Create your own learning podcasts: the generated audio features two 'speakers' who sound like typical podcast hosts, delivering content with impressive flow and informativeness. I invite you to listen to this audio overview below - you might be impressed too.
You can sign up for NotebookLM here.
One More Thing
As AI increasingly permeates everyday products, it's important to regularly review privacy and security settings in your frequently used applications.
Recently, LinkedIn quietly introduced a feature allowing the platform to use members' personal data and content for training generative AI models. This setting is enabled by default for most users, meaning their information is being utilized without explicit consent. Users in the EU, EEA, and Switzerland are not subject to this data collection, possibly due to regulations like the EU AI Act.
You can opt out of this data collection by:
Going to Settings & Privacy
Navigating to Data Privacy > Data for Generative AI Improvement
Toggling off "Use my data for training content creation AI models"
Ironically, with the backing of Microsoft and access to valuable data, LinkedIn’s AI features so far have been disappointing. I am curious how they measure their own success in their AI strategy.
I hope you found this newsletter valuable. If so, please consider sharing it with others in your network. I greatly appreciate it.
Thank you.
Joyce Li