Autostep

Cover Block

PUBLIC

Name	Autostep
Tagline	AI desktop app that maps knowledge worker time, identifies costly tasks, and automates them with AI agents. [Y Combinator]
Headquarters	San Francisco, CA, United States [LinkedIn]
Founded	2025 [Extruct AI, 2025]
Stage	Seed
Business Model	SaaS
Industry	HR / Future of Work
Technology	AI / Machine Learning
Geography	North America
Growth Profile	Venture Scale
Founding Team	Solo Founder
Funding Label	Undisclosed

Links

PUBLIC

• Website: https://autostep.ai
• LinkedIn: https://www.linkedin.com/company/autostep-ai
• Y Combinator: https://www.ycombinator.com/companies/autostep

Executive Summary

PUBLIC

Autostep is an AI desktop application that directly observes how knowledge workers spend their time, quantifying hidden operational waste and then automating it, a proposition that merits attention for its novel approach to measuring the return on investment for AI agents. The company, founded in 2025 by Aidan Pratt, positions itself as a "P&L for knowledge work," aiming to move automation decisions from guesswork to data-driven analysis [Y Combinator, retrieved 2025/2026] [HUGE Magazine, Jul 2025]. Its core differentiation lies in its passive observation method, which requires no initial integrations and builds a queryable record of operations, a contrast to survey-based process mapping or integration-heavy workflow platforms [Perplexity Sonar Pro Brief, retrieved 2024].

Pratt, a Georgia Tech student with a machine learning background, has compiled a relevant early-career portfolio, including roles at OpenAI, a BNPL lender, and a prior startup, focused on applying AI to financial and operational problems [8vc.com, retrieved 2026] [sfnet.com, retrieved 2026]. The company's early backing from Y Combinator and the venture firm Neo signals investor confidence, though the specific funding amounts and valuation are not yet public [LinkedIn]. An early proof point claims the software identified over $100,000 in operational waste within a 10-person team, a concrete ROI figure that will need validation through named customer deployments [Extruct AI, 2025].

Over the next 12-18 months, the key watchpoints will be the transition from quantified discovery to automated execution at scale, the company's ability to secure and publicize enterprise customers, and the evolution of its pricing and business model from its current SaaS structure.

Data Accuracy: YELLOW -- Core product claims and founder background are well-sourced; the key traction metric and funding details rely on limited corroboration.

Taxonomy Snapshot

Axis	Classification
Stage	Seed
Business Model	SaaS
Industry / Vertical	HR / Future of Work
Technology Type	AI / Machine Learning
Geography	North America
Growth Profile	Venture Scale
Founding Team	Solo Founder
Funding	Undisclosed

Company Overview

PUBLIC

Autostep was founded in 2025 by Aidan Pratt, who serves as its CEO [Y Combinator, retrieved 2025/2026]. The company is headquartered in San Francisco, California, and positions itself as a venture-scale startup operating in the AI-powered future of work category [Y Combinator, retrieved 2025/2026]. The founding narrative centers on the inefficiency of traditional process mapping and the difficulty of quantifying the hidden costs of knowledge work; Autostep was conceived to provide a direct-observation alternative, creating what the company calls a 'P&L for knowledge work' [Perplexity Sonar Pro Brief, retrieved 2024].

The company's early trajectory is marked by participation in prominent accelerator programs. It was part of the Y Combinator Spring batch, which the company's LinkedIn profile references as 'YC P26' [LinkedIn]. These early-stage validations from Y Combinator and Neo, the venture capital firm listed in the company's public tagline, form the core of its publicly known institutional backing [LinkedIn].

A key early milestone cited by the company is a proof-of-concept deployment where the software reportedly identified over $100,000 in operational waste within a 10-person team [Extruct AI, 2025]. This case study, while not naming the customer, serves as the primary public traction signal for the product's claimed ROI. As of the latest public information, the company is described as having between one and ten employees [Extruct AI, 2025].

Data Accuracy: YELLOW -- Foundational details confirmed by Y Combinator and LinkedIn profiles. The $100K waste claim is from a single secondary analysis; specific customer and deployment details are not independently verified.

Product and Technology

MIXED

The core proposition is an AI desktop application that maps knowledge worker activity to identify and price operational waste, then automates it. Autostep installs as a desktop client across an organization, observing real user activity to surface repetitive tasks, bottlenecks, and administrative loops that traditional process mapping often misses [Perplexity Sonar Pro Brief]. It then quantifies the cost of these inefficiencies, creating what the company calls a 'P&L for knowledge work' [Perplexity Sonar Pro Brief]. The product's initial wedge is its no-integration installation, which aims to reduce the friction of traditional workflow automation platforms [Perplexity Sonar Pro Brief]. Once high-impact tasks are identified, the system generates or recommends solutions, which the company terms 'auto-agentic' AI agents, alongside prompts, templates, and process changes [Perplexity Sonar Pro Brief]. A key technical claim is the building of a 'compounding context',a queryable record of operations that allows users to ask questions about how work is done and receive answers with supporting sources [Perplexity Sonar Pro Brief]. The platform positions itself as the 'missing scorecard for AI adoption,' measuring the return on investment for both human and automated work [X.com, 2026].

Public traction for the product is anchored on a specific, quantified outcome. According to an analysis by Extruct AI, Autostep identified over $100,000 in operational waste within a 10-person team [Extruct AI, 2025]. While the customer involved is not named, this figure serves as the primary public proof point for the product's potential ROI. The technology stack is not detailed in public materials, though the company is described as being 'built by machine learning engineers' [Extruct AI, 2025].

Data Accuracy: YELLOW -- Product claims are consistently described across multiple public sources, but the core traction metric is reported by a single analysis firm and lacks independent verification. Technical architecture details are not publicly disclosed.

Market Research

PUBLIC

The market for quantifying and automating knowledge work is moving from a theoretical exercise to a measurable operational priority, driven by the need to justify AI investments with concrete productivity returns.

Total addressable market figures for Autostep's category are not publicly available from third-party research. The company's positioning as a 'P&L for knowledge work' suggests it is targeting the broader enterprise productivity and process intelligence software market. For context, the global enterprise software market was valued at approximately $600 billion in 2024, according to Gartner [Gartner, 2024]. More specifically, the market for AI in the enterprise, which includes workflow automation and decision support, is projected to exceed $200 billion by 2028, according to a report by IDC [IDC, 2024]. These analogous markets indicate the substantial financial backdrop against which Autostep is operating, though its specific niche of desktop-observed operational intelligence remains nascent and unmeasured by major analysts.

Demand is anchored in two converging trends. First, enterprises are under pressure to demonstrate a clear return on investment from generative AI and automation initiatives, moving beyond experimentation to scaled deployment [IDC, 2024]. Second, traditional methods of process discovery, like employee surveys and manual time-tracking, are recognized as incomplete and often fail to capture the full scope of repetitive digital tasks that are prime candidates for automation [HUGE Magazine, Jul 2025]. Autostep's proposition to directly observe desktop activity aims to address this gap in visibility, providing a data-driven foundation for automation prioritization that claims to uncover waste traditional methods miss.

Key adjacent markets include Robotic Process Automation (RPA), valued at over $13 billion globally, and the broader business process management (BPM) software sector [Gartner, 2024]. These established markets represent both potential partners and substitutes. While RPA focuses on automating rule-based tasks, often at the UI layer, Autostep's approach of first mapping and costing workflows positions it as an upstream discovery and intelligence layer. The regulatory environment is currently permissive, though increased scrutiny of employee monitoring software in certain jurisdictions could introduce future compliance considerations for tools that observe desktop activity.

Data Accuracy: YELLOW -- Market sizing is based on analogous, broader categories from third-party analysts (Gartner, IDC). Specific TAM for Autostep's niche is not publicly defined. Demand drivers are supported by industry analysis and the company's own stated value proposition.

Competitive Landscape

MIXED

Autostep positions itself as a direct observer of desktop workflows, a fundamentally different approach to process optimization than the documentation-centric or integration-heavy platforms that currently dominate the market.

Company	Positioning	Stage / Funding	Notable Differentiator	Source
Autostep	AI desktop app that maps knowledge worker time, identifies costly tasks, and automates them with AI agents.	Seed; Y Combinator & Neo-backed (funding undisclosed).	Observational data capture via desktop app; generates a "P&L for knowledge work" and auto-agentic recommendations.	[Y Combinator] [Neo]
Atlassian Confluence	Enterprise wiki and collaboration platform for team documentation.	Public company (TEAM).	Deeply embedded in enterprise IT stacks; extensive ecosystem of integrations and plugins.	[Public]
Notion	All-in-one workspace for notes, docs, projects, and databases.	Late-stage private; significant venture funding.	User-friendly, flexible workspace that centralizes many knowledge functions.	[Public]
Guru	Enterprise knowledge management platform with browser extension for instant answers.	Venture-backed.	Focuses on surfacing verified company knowledge at the point of work, often in customer support contexts.	[Public]
Helpjuice	Knowledge base software for internal and external support.	Bootstrapped / private.	Specialized in creating searchable, structured FAQ and help center content.	[Public]

The competitive map for operational intelligence splits into three distinct categories. Incumbent knowledge management platforms like Confluence, Notion, Guru, and Helpjuice represent the primary substitute. Their value proposition is based on codifying processes and information into searchable documents. Autostep's wedge is that it bypasses the manual documentation step entirely, capturing how work is actually done, not how it is described. Adjacent substitutes include Robotic Process Automation (RPA) tools like UiPath and workflow automation platforms like Zapier. These require explicit integration and rule-setting, whereas Autostep aims to discover and propose automations from passive observation.

Autostep's defensible edge today rests on its proprietary observational dataset and the resulting "compounding context" it builds [Perplexity Sonar Pro Brief, retrieved 2024]. This dataset, derived from actual desktop activity across a company, is unique and not replicable by tools that rely on user-generated content or API logs. The edge is durable if the company maintains a technological lead in privacy-preserving activity analysis and agent generation. However, it is perishable if larger incumbents with distribution, such as Microsoft (through its Windows and Office telemetry) or productivity suites with deep system access, decide to build similar observational layers. The company's early backing from Y Combinator and Neo provides a talent and network advantage for recruiting in a competitive AI engineering market.

The company is most exposed on two fronts. First, its observational model faces significant privacy and security headwinds in enterprise sales cycles. Competitors with established trust and compliance certifications in regulated industries have a clear advantage. Second, while Autostep identifies waste, the actual automation step may still require integration with the very platforms it aims to displace. If an incumbent like Notion rapidly developed a lightweight, opt-in activity tracker, it could potentially replicate Autostep's discovery layer while offering a smooth path to automation within its own ecosystem, neutralizing Autostep's integration-free wedge.

The most plausible 18-month scenario is one of market definition. If Autostep can successfully convert its early beachhead into named enterprise deployments and standardize its ROI narrative, it becomes the category-defining leader for AI-driven operational intelligence. The winner in this case is Autostep, and the loser is the category of traditional, manual process consulting and mapping. Conversely, if adoption is slow due to privacy concerns or if the automation recommendations prove shallow, the winner is the incumbent knowledge management stack. Platforms like Confluence and Notion would absorb the demand for process visibility by adding more AI-powered analytics to their existing document corpus, leaving Autostep in a narrower niche.

Data Accuracy: YELLOW -- Competitor positioning is based on public company descriptions. Autostep's differentiation is confirmed by primary product descriptions, but its competitive durability against incumbents is untested.

Opportunity

PUBLIC

If Autostep successfully scales its method of quantifying and automating hidden operational waste, the prize is a new category of enterprise software that directly monetizes the multi-trillion-dollar inefficiency in global knowledge work.

The headline opportunity for Autostep is to become the default operational intelligence layer for the enterprise, a system of record for how work actually gets done. The company's core premise,that direct observation of desktop activity provides a more accurate and actionable map of process waste than surveys or manual audits,positions it to define a new category. This outcome is reachable because the initial evidence suggests the product can identify material cost savings without requiring the complex integrations that have historically stalled workflow automation projects. The claim of identifying over $100,000 in waste within a 10-person team provides a tangible, high-ROI proof point for the core value proposition [Extruct AI, 2025]. By framing its output as a "P&L for knowledge work," Autostep is speaking directly to the financial and operational leaders who control budgets for efficiency tools [Y Combinator].

Multiple paths exist for Autostep to achieve significant scale. The company's growth will likely follow one of several concrete scenarios, each with identifiable catalysts.

Scenario	What happens	Catalyst	Why it's plausible
Enterprise Land-and-Expand	Autostep is adopted as a departmental diagnostic tool within large enterprises, then expands horizontally across functions (e.g., from finance to HR to IT) as its "compounding context" builds a cross-company map of workflows.	A public case study with a named Fortune 500 company demonstrating seven-figure annual savings.	The product's desktop-based, integration-light approach reduces the friction for a single team to pilot it, which is a classic land-and-expand entry motion [Perplexity Sonar Pro Brief].
Platform for AI Agent Orchestration	Autostep evolves from a diagnostic tool into the control plane that routes identified tasks to the most effective AI agent or automation, becoming an essential middleware layer in the AI-powered enterprise stack.	The launch of an API or marketplace that connects Autostep's task-discovery engine to third-party AI agents and automation services.	The product's stated goal is not just to identify waste but to "generate or recommend AI agents" to fix it, positioning it naturally as an orchestration layer [Perplexity Sonar Pro Brief].

Compounding for Autostep hinges on its "compounding context",the queryable record of operations it builds over time [Perplexity Sonar Pro Brief]. This is not merely a data repository; it is a potential data moat. As more teams within an organization use the tool, the platform's understanding of cross-functional dependencies and enterprise-wide process bottlenecks deepens. This makes its recommendations for automation or process change more accurate and harder for a new entrant to replicate without equivalent observational data. The flywheel is straightforward: more usage yields richer context, which yields better automation recommendations, which drives more adoption and usage. While still early, the company's framing of its product as building this persistent, queryable context suggests the founders are architecting for this compounding effect from the start.

Quantifying the size of a potential win requires looking at comparable companies that monetize enterprise efficiency. Companies like UiPath, which automates repetitive digital tasks, reached a market capitalization of over $10 billion following its IPO. While Autostep operates earlier in the workflow (discovery and orchestration versus pure execution), its ambition to provide the "scorecard for AI adoption" suggests a role as fundamental as business intelligence platforms like Tableau or process mining software like Celonis [X.com, 2026]. Celonis, for example, achieved a valuation of over $13 billion by helping companies discover process inefficiencies from their system logs. Autostep's desktop-observational approach could capture a different, and potentially larger, slice of unstructured knowledge work. If the "Enterprise Land-and-Expand" scenario plays out, Autostep could plausibly target a valuation comparable to other foundational enterprise efficiency platforms. This is a scenario-based outcome, not a forecast, but it illustrates the magnitude of the opportunity if the company executes.

Data Accuracy: YELLOW -- The core product claims and founder background are well-documented, but the key traction metric (waste identified) is from a single secondary source. Growth scenarios are extrapolated from the product's stated capabilities.

Sources

PUBLIC

1. [Y Combinator] Autostep: The P&L for knowledge work | https://www.ycombinator.com/companies/autostep

2. [LinkedIn] Autostep (YC P26 & Neo) | https://www.linkedin.com/company/autostep-ai

3. [Extruct AI, 2025] Autostep Funding | Complete Analysis |

4. [Perplexity Sonar Pro Brief, retrieved 2024] Autostep Product Brief |

5. [HUGE Magazine, Jul 2025] Autostep Wants to Tell You Which Tasks to Automate Before You Waste Money Guessing | https://hugemagazine.com/feature/autostep/

6. [X.com, 2026] Autostep post on X |

7. [8vc.com, retrieved 2026] Aidan Pratt | 8VC Fellow | https://www.8vc.com/fellows/aidan-pratt

8. [sfnet.com, retrieved 2026] Aidan Pratt | Secured Finance Network | https://www.sfnet.com/detail-pages/speaker-title/aidan-pratt

9. [Gartner, 2024] Enterprise Software Market Forecast |

10. [IDC, 2024] AI in the Enterprise Market Forecast |

11. [Neo] Neo Venture Capital |

12. [Public] Atlassian Confluence |

13. [Public] Notion |

14. [Public] Guru |

15. [Public] Helpjuice |