Promote or Fire? AI Now Has an Opinion.

01 /The Concept

FrontRace was founded earlier this year. The product hasn't launched yet, but the company has already raised $4M – because the problem it's tackling is genuinely important.

The platform helps companies quickly identify which employees are performing and which aren't.

FrontRace categorizes employees into three groups:

- Thinkers: not the most active by volume or hours, but the most productive by outcomes. - Doers: work harder than anyone else, but don't produce the best results. - Non-performers: neither activity nor results – whether by disposition or by being mismatched to the role.

The method is conceptually simple. On one side, the platform analyzes all micro-activities: emails sent, messages in chat tools, participation in video calls, document edits, and similar signals. On the other side, it captures actual business results. It then maps micro-activity patterns against outcomes.

Current integrations are limited to a handful of enterprise platforms – G Suite, Microsoft 365, Zoom, and Salesforce – from which the platform pulls activity metrics automatically.

Results data comes from Salesforce, which captures pipeline movement and closed deals. Based on the initial integrations, the first target segment is clearly sales teams.

A third variable is work location: remote vs. in-office. This lets companies understand which mode is most effective for their workforce overall – and which individual employees function better with autonomy versus in-office oversight.

The output is a dashboard that ranks teams and individuals by the ratio of activity to outcomes – making it immediately clear who's at which end of the performance spectrum.

The platform also surfaces which specific behaviors and activity patterns correlate with the best results, so managers can spread those practices across lower-performing employees.

02 /Why It Matters

FrontRace's three-category model echoes what's often called the Manstein Matrix – a framework attributed to a German general who categorized military officers into four types:

- Lazy and stupid: leave them alone, they cause no harm. - Smart and hardworking: make excellent staff officers – nothing escapes their attention. - Hardworking and stupid: the most dangerous type; they generate endless unnecessary work for everyone else. - Smart and lazy: the ones to put in senior command; they get results by delegating all the "black work" to others.

A manager will eventually figure out which category each employee falls into. The problem is that "eventually" usually means too late – after significant time and money has been lost trying to get results from the wrong people. In competitive markets, the fastest win, not the smartest.

Underperforming employees also have an outsized negative effect on the rest of the team. When strong performers see that low effort is tolerated, it influences their own behavior – consciously or not.

AI's contribution here is specific: extracting non-obvious correlations from massive datasets that no human could detect manually. That's precisely what made the micro-activity-to-outcome mapping possible.

Rhythms – [covered here](/review/a-v-takom-ritme-mozhno-bolshe-zarabotat) in late last year – is working in the same space. Also pre-launch, but they raised $26M straight out of the gate.

Quan – [covered here](/review/privychka-silnee-motivacii) earlier this year – adds a sentiment layer to activity analysis. Their thesis: effectiveness and burnout prevention aren't separable, so they combine behavioral signals with employee surveys to read "morale." They've raised $4M.

Rize – [covered here](/review/mnogo-ili-jeffektivno) in February – applies micro-activity analysis at the individual level rather than the organization level. It's a personal productivity tool: an AI coach that watches what you do at your computer and helps you improve your own output. They've been building since 2022 and raised $500K in initial funding.

But individual performance is ultimately not what drives company results – team performance does. And "team" here doesn't mean a formal org chart unit; it means the actual working network of people collaborating across departments.

Confirm – [covered here](/review/rezultat-prinosit-ne-sotrudnik-a-komanda) last fall – analyzes communication-level micro-activities: email and messaging patterns between employees. From that data, it constructs the company's real interaction network. Then it evaluates the effectiveness of the informal teams that emerge from that graph – even using traditional survey instruments to add qualitative depth. They've raised $18.2M.

03 /Opportunities

Every startup mentioned here raised money or launched its product recently. The reason isn't coincidence – AI capabilities have only recently reached the quality and speed threshold needed to make micro-activity analysis reliable at scale.

But the demand has been there for years. Improving employee productivity has always mattered. In today's more financially constrained environment, it's moved from "important" to "urgent" for many companies.

The recipe itself is straightforward. Integrate with corporate platforms that record employee actions and results. Feed that into AI to find the activity-to-outcome correlation. The output:

- A segmentation of employees by the action-to-result ratio - Patterns of frequency and sequence that produce the best results per unit of effort – which can be socialized across the company

This won't deliver a 100% formula for performance improvement, but it will surface actionable insights and provide ongoing operational visibility into what's happening inside the organization.

This is a good space to move into – but now, not later. The recipe is simple enough that the next wave of startups building it is already forming. The competitive advantage goes to whoever moves fastest.

As noted above: in this race, speed beats intelligence.