Thursday, August 6
Business · Technology · Leadership

Who Are You Really Telling to Stay?

By Clark A. Ingram

Every organization has good employees and bad ones, and here’s an uncomfortable truth I’ve watched play out for more than thirty years: your day-to-day decisions are constantly sending a message about which group you want to keep. Most leaders never realize they’re sending it, and most are sending the wrong one.

The mechanism is simple. If you fail to discipline your bad employees and fail to reward your good ones, the bad ones stay and the good ones leave. You’ve told your worst people, in the only language that counts, action, that they’re welcome to stick around. Meanwhile your best people, who notice everything, hear the opposite. Over time you’re left with a roster you never would have chosen. Worse, you’ve been feeding the very turnover monster you’re trying to slay.

The good news is that the reverse is just as reliable. Discipline the poor performers and reward the strong ones, and you end up with fewer bad employees and more good ones, and some of those marginal performers will actually convert, because they finally see that good work gets rewarded. Your people already know exactly who the good and bad employees are. How you treat each group is what decides who stays.

Solving turnover is what buys you the freedom to do this. When turnover is high, you’re trapped: you keep people you know aren’t productive because you need warm bodies in the building. You can’t afford standards. Once you stabilize your workforce, you finally gain the flexibility to deal honestly with your less-productive employees instead of clinging to anyone with a pulse.

Second chances have limits, too. One company I worked with was so desperate for warm bodies that it kept rehiring the same former employees over and over. The pattern was obvious once we looked at the record: people rehired once often worked out, because they’d learned the grass wasn’t greener elsewhere, while those rehired more than once failed almost every time. So we set a simple rule, we’d rehire someone exactly once. That policy sent its own message: we’ll give you a genuine second chance, but we won’t be taken advantage of.

But I’ll warn you about what happens the moment you start making changes. Every organization has what I call THAT issue, a problem everyone knows about, everyone talks about, and management ignores. It might be the owner’s brother-in-law who’s the worst supervisor in the building and refuses to be coached. It might be favoritism, or managers who play by their own rules, or the “open door” that everyone knows is theater. THAT issue has been quietly generating turnover for years, and the monster will fight hardest to keep you from touching it, because it’s the real problem. You cannot fix your turnover while stepping around the thing actually causing it.

Underneath all of this sits a cost most leaders badly underestimate: hopelessness. The most disheartening conversations of my career were with employees I knew were excellent- people who worked late, demanded quality, trained the new hires- who had quietly concluded that nothing would ever change. “It is what it is.” That fatalism is a cancer on your culture, and it’s infinitely expensive. When one of those good people finally leaves, morale collapses and everyone else dusts off a résumé.

Here’s the encouraging part, and it’s the whole reason I do this work: hopelessness reverses fast. When you identify the real root causes, announce a plan that actually makes sense, and deliver a concrete, visible change that touches people’s everyday work, the mood shifts almost immediately. The good employees are the first to jump in; they’ve been waiting for someone to make it better. Excitement, it turns out, is as contagious as despair, and it’s infinitely more profitable. I once had an employee stop me in the hallway and whisper, “It’s getting better”, afraid she’d jinx it. That whisper is the sound of a workforce deciding to stay.

Four Ways to Send the Right Message

1. Audit the message your actions send. Look honestly at who gets rewarded and who gets tolerated. If your best people are carrying your worst and nothing changes, you’re telling the wrong group you want them to stay. Reverse it deliberately.

2. Deal with marginal performers proactively. Don’t wait for the next blowup. Sit down with the employee, agree on objective, specific behavior improvements and a timetable, and follow through. People act only when they know you’re serious.

3. Name THAT issue and fix it. Identify the known-but-ignored problem everyone in the building talks about and management avoids. Put it on the table. You cannot reduce turnover while protecting its root cause.

4. Deliver a visible early win. Announce a plan that makes sense, then produce one concrete change employees feel in their daily work. Nothing dispels hopelessness faster than proof that things are actually moving.

You are always telling someone to stay. The only question is whether it’s the people who make your company better or the people quietly driving it into the ground. Decide on purpose, and let your best people hear it loud and clear.

***

Clark A. Ingram is the Founder and President of People Profits, LLC, which focuses on the three greatest human capital problems affecting organizations: employee turnover, chronically open positions, and skills gap. He consults with a spectrum of companies and has consistently reduced turnover by more than 40 percent in the first year and achieved staffing at more than 90 percent. His new book is Churn: Proven Strategies to Overcome Failing Conventional Talent Management and Achieve Zero Turnover (People Profits, March 26, 2026). Learn more at peopleprofits.com.

TypeSafe AI Raises $40 Million for Predictable Software AI

TypeSafe AI has emerged from stealth with $40 million in seed funding led by DCVC, valuing the San Francisco startup at $200 million. Founded by former OpenAI researcher Diogo Almeida with Erik Gafni and Sasha Sheng, the company is developing AI designed to provide predictable decision-making capabilities inside software applications.

Key Takeaways

  • TypeSafe AI announced its emergence from stealth on September 15, 2026.
  • The San Francisco startup raised $40 million in seed funding led by DCVC.
  • Forbes reported the financing valued TypeSafe AI at $200 million.
  • Diogo Almeida, Erik Gafni and Sasha Sheng founded the company.
  • TypeSafe AI is developing machine-native AI intended for integration into software applications.

TypeSafe AI Emerges From Stealth With $40 Million

TypeSafe AI announced September 15 that it had emerged from stealth with a $40 million seed round led by DCVC. The San Francisco company is developing machine-native, composable artificial intelligence intended to be integrated directly into software systems.

The company was founded by Diogo Almeida, Erik Gafni and Sasha Sheng. Almeida serves as CEO and previously worked as an AI researcher at OpenAI. TypeSafe AI is building AI for software applications rather than focusing primarily on systems designed to interact with people through conversational interfaces.

The funding gives TypeSafe AI capital to develop its model technology and bring its approach to software developers. DCVC led the financing, according to the company’s announcement.

The financing follows a similar pattern seen among other AI startups developing specialized software products. AI startup Outward Intelligence’s growth, for example, has centered on using AI within a defined business application while operating with a lean team.

Forbes reported that the funding values TypeSafe AI at $200 million. The valuation was attributed to a person familiar with the deal.

The company’s launch centers on a specific technical distinction: its AI is designed to return information that software can consume directly, rather than producing conventional natural-language responses as its primary output.

DCVC Leads TypeSafe AI’s Seed Funding Round

DCVC led TypeSafe AI’s $40 million Series Seed financing. The venture capital firm described the company as a San Francisco-based startup developing models optimized for automation.

The financing supports TypeSafe AI’s development of AI that can operate as a component within software. The company’s stated approach is based on making AI outputs useful to applications that need structured information and predictable behavior.

TypeSafe AI’s launch materials describe its technology as a new class of intelligence designed to provide developers with reliable and efficient intelligence that can be integrated directly into software systems.

The distinction is relevant to the company’s product design. Conventional AI assistants generally produce responses for people to read and interpret. TypeSafe AI is instead developing models intended to return information that another software system can process.

DCVC said TypeSafe AI is building models specifically optimized for automation rather than adapting existing conversational systems to that purpose.

The company is entering a market that already includes startups adapting software products around generative AI. A separate report on AI software startup strategy changes describes companies changing products and business strategies as they incorporate AI into enterprise software.

That positioning places TypeSafe AI in the enterprise AI software market, where developers can incorporate AI capabilities into applications rather than relying on a separate conversational interface.

Diogo Almeida Leads the TypeSafe AI Founding Team

Almeida founded TypeSafe AI with Gafni and Sheng after working on artificial intelligence research at OpenAI. Forbes reported that Almeida spent four years at OpenAI working on improvements to ChatGPT’s responses before leaving the company in 2024 to start TypeSafe AI.

Almeida’s previous work is directly connected to the company’s approach to AI models. The company is focused on building a system that can provide software with structured intelligence and information about the confidence of its outputs.

TypeSafe AI’s founders are developing the company around the distinction between AI built for human interaction and AI built for machine consumption. Its technology is intended to let software use model outputs as inputs for subsequent actions or decisions.

The company has described this approach as machine-native and composable. Its public announcement says the goal is to give developers intelligence that can be integrated directly into software systems.

The founding team’s focus also addresses a specific limitation identified by TypeSafe AI: AI systems that work well as assistants can produce outputs that are less predictable when incorporated into automated software workflows.

The company’s founding model has similarities to other specialized AI startups that have raised capital around defined business applications. Cascade’s AI construction platform, for instance, applies predictive data analysis to a specific industry workflow rather than developing a general-purpose consumer AI product.

TypeSafe AI Develops Machine-Native Software Intelligence

TypeSafe AI’s model is designed to return numerical responses along with probability estimates and scores indicating confidence in individual outputs. The information is intended to be consumed by software rather than presented primarily as conversational text.

That structure is designed to give applications information about both an AI-generated decision and the model’s confidence in that decision. Software can then use those values when determining whether an action should be automated or receive additional human review.

Insurance underwriting is one example of the model’s intended use. A system could examine evidence related to a property and return a probability associated with whether there was a history of fire at the property.

TypeSafe AI Raises $40 Million for Predictable Software AI

Photo Credit: Unsplash.com

The approach differs from a conventional language-model response because the output is intended to provide a structured decision signal rather than a paragraph that a person must interpret.

TypeSafe AI says its first model is designed for direct integration into software systems. The company describes its technology as composable AI, meaning developers can use the model as a component within larger applications.

The company has also claimed that avoiding text generation allows its models to operate at lower cost and higher speed than conventional frontier models. Those performance figures remain company claims.

The focus on software-native AI also parallels other efforts to build artificial intelligence directly into development workflows. 8090 Labs’ enterprise AI software platform is another example of a startup developing AI around software production rather than treating AI solely as a standalone conversational product.

The Startup Targets Predictable AI Decision-Making

TypeSafe AI’s product strategy centers on making AI outputs easier for software to evaluate before taking action. Its model provides probability estimates and confidence scores that can be incorporated into application logic.

For software developers, that structure creates a defined distinction between an AI output that meets a specified confidence threshold and one that requires additional review. The company’s stated objective is to support automation in cases where the software can determine that an output is sufficiently reliable.

TypeSafe AI’s technology is also designed around the idea that software needs different AI characteristics from a conversational assistant. Its announcement says existing frontier models can hallucinate, change their methods between requests and introduce variability into systems that require predictable behavior.

The company’s model therefore focuses on outputs that applications can process repeatedly. Rather than requiring a person to interpret each response, the system is intended to provide structured information that can become part of a software workflow.

TypeSafe AI emerged from stealth with its first public model, Jev, according to the company’s launch announcement. The model is positioned as machine-native intelligence intended to operate directly inside software systems.

The company’s $40 million seed financing will support the development of that technology as TypeSafe AI builds its business around AI designed for software-based decision-making.

Frequently Asked Questions

What is TypeSafe AI?

TypeSafe AI is a San Francisco-based AI startup founded by Diogo Almeida, Erik Gafni and Sasha Sheng. The company develops machine-native AI designed for direct integration into software applications.

How much funding has TypeSafe AI raised?

TypeSafe AI emerged from stealth with a $40 million seed round led by DCVC. The financing valued the company at $200 million, according to Forbes.

Who founded TypeSafe AI?

TypeSafe AI was founded by Diogo Almeida, Erik Gafni and Sasha Sheng. Almeida is the company’s CEO and previously worked as an AI researcher at OpenAI.

What does TypeSafe AI develop?

TypeSafe AI develops AI designed for direct use inside software systems. Its approach includes structured outputs, probability estimates and confidence scores intended to help applications process AI-generated decisions.

Where is TypeSafe AI based?

TypeSafe AI is based in San Francisco. The company emerged from stealth with its $40 million seed financing on September 15, 2026.