5 Hype Cutting Truths About AI

The hype wave rolls on and as usual statements and conjecture have become a distraction.

The AI revolution continues. Each week this spring and summer seems to bring a major AI announcement or shocking news story. With the hype-driven news cycle has come increased interest and fears about AI agents, workforce replacement, and significant societal impacts.

Certainly click worthy for most, the ongoing dialogue, at least presented by AI companies, media, and influencers, offers more of a distraction than help regarding AI adoption. Here are what I perceive to be the five most significant truths about the current AI wave that many of us in the business discuss amongst ourselves

1) A Longer Adoption Horizon than Anticipated

The AI workflow “replacement” horizon is longer than many companies, media, and AI influencers are discussing. The reality is that most agentic frameworks and technologies are still not ready for prime time with piss-poor results usually under 50 percent success rates.

Tech layoffs targeting staff with agents are, in most cases, premature and experimental. Many lack clarity on timeframes and the type of staff lay-offs. Companies often rehire some of their staff when they realize the negative impacts of immature AI unleashed on their customers.

These announcements are also performative announcements for investors and customers alike. Specifically, tech companies in the AI business need to eat their own dog food, and as publicly traded companies, need to lead the efficiency wave created by so-called agents.

However, most agentic frameworks underperform their human counterparts. Even Salesforce, creator of the much-touted Agentforce, acknowledges this with its recent research study on LLM-based agentic AI. The study found, “leading LLM agents achieve approximately solely 58% single-turn success rate on CRMArena-Pro [19 finely tuned Agentforce tasks], with significant performance drops in multi-turn settings to 35%.”

LLM weaknesses must be addressed before agents can be fully trusted in customer service. Unfortunately, my wife found this out after communicating with Delta Airline’s customer service agent this week.

2) Wild Statements About Job Loss Are Outliers

We’re seeing a lot of conjecture resulting from AI-centric layoff announcements, which seem to be dropping about once a week. In addition, job loss fears went into overdrive after Anthropic CEO Dario Aodi’s appearance on CNN last month, declaring that 50% of entry-level knowledge working jobs and 20% of the jobs are endangered due to AI. Aodi’s comments follow OpenAI Altman’s 2024 remarks that about 95% of marketing jobs will likely become displaced by AI.

These boasts and warnings claiming widespread depression-level unemployment and displacements certainly make for great headlines, but they lack substantive data behind them. In a conversation on the No Brainer podcast with Brent Orell, American Enterprise Institute Senior Fellow, called these statements from AI companies outliers. This podcast drops tomorrow.

Outlier statements are a PR ploy destined to accomplish two things. Media visibility and a sense of omnipotence about AI software companies’ technological capabilities. While, their worst-case scenarios may be untrue, certainly, their technologies are worth a spin. Nothing works like marketing through fear.

3) Most AI Influencers Are Just Quote Farmers

As you can already tell, many bold statements are being made. Rather than questioning them and analyzing their veracity, the media and AI influencers would rather farm quotes and aggregate such statements.

Case in point, yesterday’s TechRadar coverage of Agentforce 3.0, which parrots the announcement video — literally a fake newscast announcing the features and use cases (you can’t make this shit up)— with quotes. Why aren’t people questioning the veracity of Agentforce? There’s a lot to be wary about:

• The above-cited Salesforce research.

• Slow Agentforce sales, which may have necessitated the Informatica acquisition.

• The third iteration of Agentforce finally adds a management layer for enterprises so they can actually measure agentic results, but it does not address the aforementioned LLM weaknesses.

• And then there is the murky, car dealership-like sales process with Salesforce around agents. It's expensive and unclear whether you will actually get results.

Most AI influencers seem to regurgitate the latest AI release or statement with an added level of spin, and a CTA to buy their latest offering. It’s good marketing (in the short term), but it’s not valuable news.

However, in an era when even the most pristine institutions have been compromised to serve ulterior motives, there is little recourse for those who trust others for their news. Given how far digital media literacy has declined and the general public’s over-reliance on algorithmically driven social media-based clickbait, this situation is unlikely to change.

In the long term, trust in companies, media, and AI influencers will continue declining, which will only create more friction for enterprise adoption.

4) People Are Paralyzed

As technology companies create hype, they are not speaking to customers. Technology capability does not translate to specific work problems and results.

While adoption and attempts to operationalize AI agents increase, most businesses still struggle to incorporate basic AI functionality into their operations. They have broken data infrastructure and operations, an uncertain economic environment that’s threatening profitability, and a lack of clarity about how AI benefits their organization.

Meanwhile, most employers try to address AI adoption problems by demanding that new employees have AI skills. As noted last week, they’d rather hire AI-capable workers than upskill. This is creating immense demand on the workforce to embrace AI tools independently and demonstrate their capabilities. Yet, many are uncertain of how or where to start their efforts.

Unfortunately, most enterprises are in for a rude awakening. Staff skills are a start, but AI will at best create marginal improvements until organizations address broken tech stacks, data governance behaviors, cultures, and business processes. In short, retail therapy in the form of AI tech and staff doesn’t solve their problems.

5) Most Enterprise-Level AI Adoption Will Come Via Existing Platform Upgrades

Given the friction businesses face when acquiring AI talent and tech, not to mention overcoming their own internal challenges, the safe bet is to wait for their existing AI platform vendors to incorporate AI. Every forward-thinking tech vendor is responding to the AI revolution and incorporating capabilities into their new releases.

Some of them are pretty useful, including:

• Hubspot’s continued AI evolutions and integrations to allow for a more nimble CRM

• Adobe’s answer engine optimization tools

• Zapier’s MCP usage to tie AI to existing automation

• SAP’s analytics and task automation in its ERP platform

Small and medium enterprises, in particular, will look to their existing vendors rather than incur the risk of adopting an AI-specific platform or perhaps custom implementations. It’s just easier.

While that may not be an ideal technology solution and may be slower, it is the one that allows them to adopt with the least disruption and lowest risk. And that, for most companies, is good enough.

Certainly, some existing software platforms—particularly those that don’t integrate or acquire their own AI solutions—will be disrupted by newer AI-centric startups. But this is also part and parcel of normal technology evolution. New, better solutions arise and win. AI is just the latest lubricant in the innovation cycle.

So, these are my five hype-cutting truths about AI. What do you think? Are they on point? What would you add?

The last three images were created on Midjourney.