Benedict Evans on AI, Tools, and Enterprise Transformation

AI Won’t Instantly Turn Everyone Into a Tool‑Builder

Takeaway: AI will not make software development trivial for every employee, nor will it instantly eliminate the complex, layered software stacks that large enterprises rely on. The real impact lies in how AI expands existing tools, creates new options, and forces companies to rethink automation at scale.

1. The Myth of Universal Tool‑Building

Common belief: Large language models let anyone ask a model to write code, turning every worker into a developer.
Reality: Most workers never think about redesigning their jobs. Professionals such as lawyers or salespeople focus on domain expertise, not on building software.
Evidence from Evans: He notes that products like Excel provide templates and assistants, but each template ultimately becomes a separate product—illustrating that AI‑generated tools are not a universal solution.

"If you spend all your time in the Silicon Valley bubble, it can be easy to forget this, because your entire world is about creating tools that change how things are done." – Benedict Evans

2. Enterprise Software Landscape Today

Complex stacks: Companies run dozens of vertical SaaS apps, massive ERP systems (SAP, Workday), and countless spreadsheets, scripts, and ad‑hoc tools.
Visibility problem: Organizations often lack a clear inventory of what software they actually use and pay for.
Automation gap: Even with this software, many repetitive tasks remain unautomated because the need for a tool is not obvious.

3. The Role of the “Forward‑Deployed Engineer”

Definition: An engineer who can spot automation opportunities that domain experts overlook.
Why needed: The problem often lies in recognizing a need for a tool, not in building it.
Commentary: Users may not see the inefficiency; a technically savvy observer can propose AI‑driven solutions.

4. Institutional vs. Improvised Solutions

Institutionalized tasks: Handled by established systems (SAP, Workday) where consistency, audit, and security are required.
Improvised tasks: Managed with free‑form tools (Excel, email, shared folders) for edge cases and one‑offs.
Transition point: When an improvised process scales, companies must institutionalize it, adding governance and compliance.

5. How AI Alters the Existing Spectrum

Extension of current tools: AI adds capabilities to Excel, Tableau, email, and other improvisation layers.
New free‑form spaces: Chatbots become another improvisational layer that can both take over tasks and hand them back to traditional apps.
Threshold shift: AI does not remove the decision of what to automate; it merely changes the cost and speed of building those automations.

6. Real‑World Enterprise Adoption Patterns

Pilot‑centric approach: Companies run AI pilots; roughly half succeed, mirroring historic technology rollouts.
Usage distribution: A small minority become power users, a moderate group uses AI occasionally, and the majority sees little impact.
Change‑management challenge: Similar to the rollout of PCs in the 1980s or browsers in the 1990s—technology diffusion is uneven and requires training and governance.

7. Strategic Questions for Executives

Acquisition strategy: Should the firm buy, build, or pilot AI solutions? Which vendors or internal teams should lead?
Operational impact: How does AI reshape existing workflows (e.g., email, spreadsheets) for specific industries?
Economic & competitive implications: Does AI create new revenue streams, cost pressures, or existential threats?

8. The Human Element Remains Crucial

Accountability & audit: AI cannot replace the need for human oversight, especially for security, durability, and regulatory compliance.
Domain expertise: Successful AI tools require deep knowledge of the business problem, not just code generation.
Commentary highlights:
- zacksiri argues AI will flatten software hierarchies, but this overlooks governance needs.
- ryuuseijin stresses that accountability cannot be outsourced to AI and highlights alignment risks.
- iLoveOncall shares a cautionary tale where a non‑engineer’s AI‑generated script caused hidden errors, underscoring the need for expert review.

9. Outlook

Short term: AI will augment existing tools, enabling faster, low‑code automations for well‑defined tasks.
Medium term: Companies will face a wave of new SaaS offerings that blend AI with traditional vertical applications.
Long term: The transformative value of AI will emerge from capabilities that were previously impossible—not merely from speeding up current workflows.

Conclusion: AI reshapes the software ecosystem by expanding the capabilities of both institutional and improvised tools, but it does not eliminate the fundamental challenges of identifying automation opportunities, securing governance, and achieving enterprise‑wide adoption. Executives must ask the right strategic questions and involve domain experts to harness AI’s true transformative potential.