{"path":"/beyond-the-application/applying-intelligence-late/agents-as-software-discoverers","content":{"title":"Agents as Software Discoverers","summary":"Agents may increasingly recommend, evaluate and select software on behalf of their users.","body":"Today, software discovery is largely a human activity. People search, read reviews, compare features, trial products, learn interfaces and work out integrations.\n\nAn agent with context about the user's goal could do much of this on their behalf. It might encounter a problem, look for a suitable capability, inspect its tools, compare cost and reliability, and recommend or use it.\n\nThat shifts discovery from **a separate shopping activity** to **a step in getting work done**.\n\n## Why this matters for small products\n\nA highly specialised service may have trouble reaching customers. Its potential users do not know it exists, and the problem is too small to justify a lengthy search or integration project.\n\nAn agent can discover the service at the moment its value is obvious. This could make [software slivers](/beyond-the-application/the-disaggregation-of-software/software-slivers) economically viable and reduce the advantage of having a feature already bundled inside a popular application.\n\nAgent-friendly descriptions, discoverable operations and clear pricing become part of the distribution strategy.\n\n## A new intermediary\n\nThe discovery agent could also become a gatekeeper. How does it rank options? Whose interests does it serve? Can the user inspect why it chose a product, what permissions were granted and what will be paid?\n\nTrustworthy discovery should allow comparison, reversibility and user control, particularly for consequential purchases or access to private data.\n\nThe claim is directional, not that every software purchase will become automatic. Users will still sometimes want to choose a familiar vendor or maintain a stable workflow.\n\nThis is one mechanism behind [Applying Intelligence Late](/beyond-the-application/applying-intelligence-late): the selection decision is made when the need is known, rather than entirely during the original product-design process."}}