A DIY content library for LinkedIn repurposing
The tools to index a LinkedIn back catalogue are now a weekend project. Whether B2B teams treat their archive as inventory is a separate question.
Key takeaways
- A Buffer developer built a searchable, AI-queryable library of her LinkedIn posts using the public API and an LLM.
- The technical barrier to treating a LinkedIn archive as a structured dataset has effectively disappeared.
- Most B2B comms teams cannot answer which past posts drove meaningful engagement from named buyer segments.
- The compounding asset on LinkedIn is the back catalogue, not the next post.
- Regulated organisations gain a governance dividend: pre-cleared language patterns speed up legal review.
Shivani, a developer at Buffer, built a personal content library on top of the company's own API: a searchable archive of her past LinkedIn posts, an AI chat layer for analysing what worked, and a two-way loop that pushes drafts back into Buffer's Create space. The Buffer Resources write-up reads as a builder's diary. It is also, quietly, a verdict on how most B2B teams treat their LinkedIn output: as exhaust rather than inventory.
The mechanics matter. Posts sync from Buffer into a Postgres database. A retrieval layer lets her query by topic, format, or engagement. An LLM sits on top to surface patterns ("which of my hooks beat the average", "what did I say about onboarding in March"). Approved rewrites flow back into Buffer as scheduled drafts. The whole thing is one developer's weekend project, not a SaaS launch.
That is the point. The tools required to treat a LinkedIn feed as a structured corpus, rather than a stream that vanishes after 72 hours, are now within reach of any competent in-house team. Buffer exposes the API. OpenAI and Anthropic expose the models. A junior engineer can wire them together in an afternoon. The barrier is no longer technical. It is whether the marketing function has decided its own archive is worth indexing.
Most have not. Walk into a financial-services comms team and ask which of the CEO's 200 posts last year produced the most inbound from buy-side analysts, and you will get a shrug or a screenshot. Ask a multilateral which framing of climate finance drew the most qualified comments from member-state officials, and the answer lives in someone's head, if anywhere. The data exists. LinkedIn's native analytics expose it. Nobody is pulling it into a queryable shape.
This is the gap Shivani's project illustrates. The compounding asset on LinkedIn is not the next post. It is the back catalogue: the hooks that earned saves, the carousels that drove profile visits, the comment threads where a partner at a competitor showed up. Treated as a database, that history becomes a brief for the next quarter's content. Treated as a feed, it becomes nothing.
The repurposing loop is where the productivity case lands. A communications lead at an industrial group running ten executive accounts spends most of their week drafting from scratch. A library of past winners, tagged by theme and ranked by meaningful engagement, collapses that work. The same opening line that performed for the CFO in May can be adapted for the COO in November, with the LLM doing the first pass and a human doing the edit. The output is not more posts. It is fewer, better ones, drawn from evidence rather than instinct.
There is a governance dividend too. Policy institutions and UN agencies that vet every executive post through legal and comms could, with a structured archive, pre-clear language patterns rather than re-litigating each draft. "We have said this before, in this form, and it cleared review" is a faster path to publication than starting from a blank page. The library becomes an institutional memory that survives the staff turnover these organisations are famous for.
The catch is discipline. A content library is only useful if someone keeps the tags clean, the engagement data fresh, and the AI prompts honest. Left to rot, it becomes another dashboard nobody opens. The Buffer post is candid about this: Shivani built it for herself, uses it weekly, and iterates. Enterprise teams that commission a similar tool and then assign it to no one will get the result they deserve.
The wider signal is that the creator-tooling layer around LinkedIn is bifurcating. On one side, the all-in-one schedulers competing on UI polish. On the other, APIs and LLMs that let serious teams build exactly the workflow they need. The interesting B2B brands over the next year will be the ones that stop buying generic tools and start treating their LinkedIn presence as a proprietary dataset. Shivani's weekend project is a preview of what an in-house version looks like. The question for heads of communications is whether their own archive is worth the same weekend.