All Roads Lead to Rome
A Primer on Different Approaches to Autonomous Software Development with AI
AI is transforming software development, one of the most critical tasks in technology. Companies are racing to achieve fully automated software development systems. Imagine a world where anything you dream up can be developed instantly—it’s modern-day magic. I’m on this journey too, and I want to share what I’ve learned and observed about the different paths to making this vision a reality.
The Ultimate goal: Bespoke software at scale
We use software in every aspect of our lives, each with unique contexts and challenges. Imagine solutions materializing instantly at the mere mention of a need. This isn’t just a fantasy; it’s the promise of autonomous software development.
Bespoke solutions shine because they address specific needs with pinpoint accuracy, delivering the most effective results. In the past, only tech giants like Google or Facebook could afford custom software due to high development and maintenance costs. Now, with AI-driven automated software development, this capability is within everyone’s reach.
Autonomous software development doesn’t just handle mundane tasks; it tackles complex problems and tailors solutions to individual requirements. This democratization of bespoke software will transform our world, and whoever achieves this first will define the future.
The Roads
Three distinct approaches emerged from analyzing hundreds of projects and companies. Each has unique strengths and challenges. I’ll break down why they matter, show real-world examples, and explore their current status and future directions—including go-to-market strategies, target users, pricing models, and the key leverage points or potential challenges that could shape their success.
Copilot
The Copilot-style products assist developers during their work. Typically, they are IDE extensions or heavily modified IDEs, acting like a modern-day 'Clippy' that actually provides value. Some take a more native approach, integrating deeply with existing IDEs to offer a more natural experience when interacting with AI. This route requires low commitment to start using, and users have full control over its use.
Why This Approach?
Given that current AI technology isn't yet capable of full automation, companies adopting this approach aim to deliver immediate value and build trust over time. By integrating into developers' existing workflows, they can accumulate valuable data critical for achieving full automation in the future. Even if full automation isn't achieved, they have a product that works and is loved by users, ensuring ongoing relevance.
Example Product
GitHub Copilot, Cursor, Codeium, AugmentCode
These tools assist developers during their work, typically as IDE extensions or modifications. They provide intelligent code suggestions, chat assistance within the IDE, and advanced code search capabilities. By integrating seamlessly into developers' existing workflows, they enhance productivity and reduce friction in the coding process.
Product Capabilities Now
Currently, these products offer features like code completion, providing intelligent suggestions as developers type. They include chat assistance within the IDE, allowing developers to ask questions and receive explanations without leaving their coding environment. Additionally, they offer advanced code search capabilities to help developers quickly find code snippets or references within large codebases.
Next Product Capabilities
The next steps involve expanding capabilities from single-file actions to multi-file actions, enabling changes that span across multiple files and modules. There's also a focus on enhancing existing features, such as improving documentation generation, writing test code, and offering more sophisticated refactoring and debugging tools. Many companies are likely working on secret projects to develop next-generation products involving autonomous agents, aiming to stay ahead in the evolving market.
GTM
The go-to-market strategy typically involves product-led growth. Companies offer free trials or freemium versions to encourage adoption, upselling additional resources or premium features later on. This low-friction model allows developers to start using the tool with minimal commitment, fostering organic growth within the developer community.
Target Product User
The primary users are developers—individuals actively writing and maintaining code. These tools are designed to integrate seamlessly into their daily workflows, providing assistance without imposing significant changes to their existing processes.
Target Buyer
While developers are the primary users, the target buyers can also include businesses and organizations. Sometimes, companies purchase licenses for their teams to standardize tools and improve overall productivity. However, the initial focus is often on developers, with business adoption following once the tool’s value is demonstrated.
Pricing
Pricing models are usually subscription-based per seat, typically ranging from $10 to $20 per user per month. Tiered usage limits or additional features at higher pricing tiers allow companies to cater to both individual developers and larger teams.
Next Strategic Move
Strategically, these companies aim to take on more responsibilities from developers by automating additional aspects of coding, thereby increasing reliance on the tool. They plan to leverage the trust and brand recognition they've built to introduce more advanced features. Continuing with product-led growth maximizes marketing efficiency and maintains user adoption momentum.
Leverage
The key leverage points include the trust built over time as users experience consistent value from the tool. Established brand awareness gives them a competitive edge in attracting new users. Their efficient growth model, centered on product-led strategies, reduces customer acquisition costs and fosters a loyal user base.
Drag
There's a risk of designing too much for the present and not adequately preparing for future advancements, which could lead to being overtaken by more forward-thinking solutions. Relying heavily on current success may breed complacency, hindering innovation in a rapidly evolving field.
“Out of the Gate” Autonomous Agent
This approach involves products designed to perform any task that a typical human developer can handle. Users provide a problem statement, and the AI builds the solution autonomously. From the beginning, these companies aim to create fully autonomous agents without relying on gradual improvements.
Why This Approach?
Instead of building a long road toward the ultimate goal, these companies are heading straight for it. They want to have the product that's designed for the future, without being bogged down by the limitations of current technology. By focusing on where the industry is going, they hope to leapfrog incremental advancements and deliver groundbreaking solutions.
Examples
Devin, Factory, Poolside.ai, codegen.com, second.dev
These companies are developing platforms where users provide problem statements, and the AI autonomously builds the solution, aiming to redefine software development by eliminating the need for manual coding.
Product Capability Now
Currently, these products use graphical user interfaces (GUIs) to interact with users. Users engage with the agent during problem statements and general Q&As, often using a chat-based UI to provide additional directions. They do not interact with the code generation process directly, making it a very passive experience. Due to current limitations of AI models, these products usually include an 'escape hatch' that allows users to interact with the code directly if the AI agent cannot complete the task.
Next Product Capability
Improving the developer experience is crucial; enhancing usability and reliability is key to gaining user confidence. When developers need to step in, any friction or poor experience can significantly undermine trust in the product. They need to refine how humans interact with the AI agents, ensuring that any required intervention is seamless and intuitive.
Target Product User
The users are typically engineers who manage pull requests and repositories. Unlike Copilot-style products, the user profile here is less defined. Engineering managers and other team members who want to complete projects without heavily relying on developers may also find value in these tools.
Target Buyer
The focus is on businesses rather than individual developers. These products are often marketed to organizations looking to streamline development processes and reduce dependency on human engineers.
GTM
Their strategy usually involves demos and waitlist gating. By showcasing the product's capabilities through demonstrations, they generate interest and manage access through waitlists to create exclusivity and anticipation.
Pricing
There's limited information on pricing models for these products. They need to balance affordability for individual users with scalability for enterprise clients. For example, considering Cursor's pricing—$20 per month for 500 premium requests—that's about $0.04 per request. If one request equates to generating a function and a project requires between 300 to 30,000 functions, costs could range from $12 to $1,200 per iteration. These figures do not directly translate to the actual prices users pay for 'Out of the Gate Autonomous' products.
However, companies adopting the 'Out of the Gate Autonomous' approach are likely using sales-driven pricing models. Pricing is typically contract-based and requires negotiation, tailored to the specific needs of each client. This approach allows them to offer flexible solutions that scale with different organizations. While this provides customization and value alignment, the lack of upfront pricing transparency can be a barrier, as it necessitates engaging with a sales team before understanding the cost.
Next Strategic Move
The immediate goal is to demonstrate the product's ability to effectively perform development tasks. They need to put the product into customers' hands and expand into new engineering domains. Building credibility through successful implementations is essential for gaining trust and attracting a broader user base.
Leverage
They have the advantage of being inspirational, capturing the imagination of innovators and early adopters. By building where the industry is heading, they position themselves at the forefront of technology. This visionary approach can attract attention and investment from those excited about the future of AI-driven development.
Drag
However, they face significant challenges. They need to deliver a value-producing product quickly to maintain interest and gain traction. The lack of a clearly defined target user profile can hinder marketing and product development efforts. Balancing the need for accuracy and speed is difficult; delivering precise solutions rapidly is essential for user trust.
There's also the challenge of balancing big dreams—creating generalized solutions—with the need to provide immediate value through more focused applications. Convincing customers to adopt a radically new approach requires them to take a leap of faith, making the sales process harder.
Service Augment Approach
Rather than trying to convince users of value upfront, this approach delivers value first and introduces the platform later. Companies using this method provide services that directly solve customer problems, effectively demonstrating their capabilities before offering a product. By focusing on tangible results, they build trust through proven outcomes.
Why this approach?
Customers demand real, impactful outcomes and don't want to wait for incremental improvements; they seek immediate solutions to pressing needs. By delivering results upfront, these companies meet expectations head-on. Focusing on one job exceptionally well allows them to establish credibility and a strong reputation. Once they've proven their effectiveness in a specific area, they can expand offerings and enter new markets with the trust they've built.
Examples
Isoform, Mechanical Orchard, Grit.io
These companies offer services that directly solve customer problems by delivering final business results, such as pull requests, new repositories, or completed projects. By focusing on delivering tangible outcomes before introducing a platform, they build trust and demonstrate their capabilities, effectively addressing specific needs with precision.
Product Capability Now
Currently, these companies deliver final business results-such as pull requests, new repositories, or completed projects-by leveraging AI under the hood. They tailor their services to different domains, industries, and company sizes, addressing specific needs with precision. By focusing on delivering concrete outcomes, they establish credibility and showcase the effectiveness of their AI-driven solutions.
Next Product Capability
Their next steps involve expanding into new verticals and offering adjacent value. By broadening their industry focus or problem types, they can reach a wider audience and unlock growth opportunities. Transitioning from a service model to providing a platform or product is essential for scaling and reducing operational complexities.
Target Product User
In this approach, there isn't a direct end-user in the traditional sense. Similar to a development shop, they aren't providing tools for users to interact with but are delivering completed solutions. The focus is on fulfilling the client's needs without requiring the client to engage with a product interface.
Target Buyer
The target buyers are CTOs and engineering leaders within organizations. These decision-makers have the budget to spend on external engineering resources to get projects done efficiently. They're interested in solutions that can be delivered quickly and effectively without overburdening their internal teams.
GTM
Their go-to-market strategy is sales-led. They leverage networks, customer testimonials, and cold outreach to generate interest. Using 'contact us' prompts on their websites, they capture leads from interested visitors. They can also use channel sales, partnering with other companies or intermediaries to reach a broader customer base. This multi-faceted approach builds relationships and moves prospects through the sales funnel.
Pricing
Pricing focuses on outcomes to establish trust. Instead of flat fees or subscriptions, they base pricing on the value delivered. This aligns their interests with the client's success, fostering a partnership mentality. In my previous article on B2B pricing in the AI era, I discussed how emphasizing value and outcomes builds stronger customer relationships. By pricing based on results, they reduce client risk and demonstrate confidence in their ability to deliver.
Next Strategic Move
Their next steps involve expanding into new verticals, offering adjacent value, and providing more sustainable solutions. By broadening their industry focus or the types of problems they solve, they can reach a wider audience and unlock growth opportunities. Offering sustainable value ensures long-term client relationships and recurring revenue streams. Transitioning from a service model to providing a scalable product or platform is essential for growth and reducing operational complexities.
Leverage
They have a quick path to resolution, which helps in gaining trust rapidly. By delivering tangible results promptly, they prove their value and build strong client relationships. With a clear target buyer and an established sales process, there's less friction in customer acquisition. Their focus on specific buyer personas enables them to tailor their messaging and approach effectively.
Drag
However, there's a race between product maturity and operational complexity. The team must advance the product quickly to avoid becoming bogged down by service delivery. While useful initially, the service approach can become a crutch. To scale and stay competitive, they need to transition away from services to product development. This shift requires courage and decisive action, involving redefining business models and potentially disrupting revenue streams.
Wrapping Up
Choosing the right path toward autonomous software development depends on your context and resources. Each approach—Copilot Style, Out of the Gate Autonomous, and Service Augment—has unique strengths and challenges. Aligning your strategy with your goals, capabilities, and customer needs is essential. As you build trust and gather valuable data, you'll move closer to delivering bespoke software solutions at scale. The journey to 'Rome' may differ for each of us, but with determination and the right approach, you'll be among the first to arrive.