What the Library Got

Ellen Cho is preparing a grant application for a state technology infrastructure grant, and she has a question.

The question is about a community partnership structure that Howard Brennan recommended nine months ago, before he returned to Cincinnati, before his Native Sonia Park left for a new position in Seattle, before the deployment formally ended. The strategic plan Howard built is on Ellen’s desk. The community partnership framework Sonia documented is in the shared drive. The knowledge library the AI captured across six months of deployment is available through the system. Ellen’s question is about why a specific partnership structure was chosen over an alternative that appeared in the notes but was not selected.

She asks the knowledge library.

The library returns Howard’s reasoning from session four. The explanation is clear and complete. It covers the institutional context, the political considerations, the resource requirements, and the specific reason the alternative structure was not feasible given the library’s current staffing model.

Ellen reads it. She understands the words. She reads it again. She is not sure she understands the reasoning behind the reasoning: the judgment about the city’s funding landscape that Howard brought to the decision, the specific reading of the county library board’s political dynamics that shaped the analysis, the institutional knowledge accumulated over thirty-four years of running library systems that is present in the explanation but not fully visible in it.

She submits the section using Howard’s framework. It is the right framework for the grant. She does not know if she is applying it the way Howard would apply it to this specific grant context.

This is what the knowledge library does and does not hold. The record is accurate. The judgment behind the record is only partially present.

Howard Brennan spent thirty-four years building library systems in Ohio, twenty-two of them as a director. He knows how a public library’s strategic planning process works, how it fails, and how to distinguish the two from inside the institution’s politics. He knows how community partnership frameworks are funded and which partnerships produce value and which produce friction without return. He knows how a library board reads a strategic plan and which elements of a plan will generate resistance from a board member who has a prior position on technology investment. He knows all of this the way people know things they have never had to articulate: through experience accumulated over decades in specific institutions with specific political contexts.

When BGO placed him at the Ohio library system, he came with the map. Sonia Park, 26, just out of her MLIS program at the University of Michigan, came with the ability to make the map readable. Sonia’s facility with community needs data, her speed with documentation tools, and her instinct for what a grant application committee would need to see were what carried Howard’s strategic thinking from his head into a form the library could use. The deployment ran six months. The strategic plan, the collection policy restructuring, and the community partnership framework were the formal deliverables.

Ellen Cho, who had been director for four years before the deployment, received all of them. She did not receive Howard.

The strategic plan transferred. That sentence should be said specifically before noting its limits.

The five-year plan Howard developed with Sonia’s analytical support is a document the library is using as its operational guide. The prioritization criteria are clear. The metrics for assessing progress are built in. The grant the library is currently applying for is one of the four funding priorities the plan identified. The plan is not theoretical. It is the library’s working map.

The collection policy restructuring transferred. The methodology Sonia developed for community needs assessment, drawing on demographic data, library usage patterns, and gap analysis from peer institutions, is a tool the library staff is now using independently. They ran the methodology themselves for the grant application. They did not need to call Sonia.

The community partnership framework transferred in a specific and limited way. Eight partnerships were identified, approached, and brought to varying stages of active relationship during the deployment. Six of those partnerships are active nine months later. Howard built them. The staff is maintaining them. Maintaining an existing relationship is different from building a new one from the strategic analysis Howard used to identify the eight in the first place.

These are durable deliverables. They are doing what they were built to do. They are not the same as Howard.

The knowledge library holds what can be structured.

Howard’s reasoning for the strategic plan’s prioritization choices is in the library, and it is useful. When Ellen needs to explain to a board member why the technology infrastructure grant was ranked above the literacy program expansion in the strategic plan, the library gives her Howard’s reasoning: the assessment of the state grant landscape, the specific funding climate for library technology versus programming, the capacity analysis showing the technology investment as a prerequisite for the programming expansion rather than a competitor to it. The library gives her all of this. She can use it.

The methodology Howard used for his community needs assessment is in the library. The approach to mapping funding sources against partnership capacity, the criteria for distinguishing a sustainable partnership from a transactional one, the framework for reading institutional political dynamics that Howard applied to the library’s city council relationships: all of this is documented, structured, and queryable.

Ellen accesses these records regularly. The library has answered 41 of her 43 queries in a way she considers useful, a ratio that aligns with Patricia Stone’s experience at the West Virginia health center.

The other two required Howard’s email address.

The knowledge library does not hold what cannot be structured.

The judgment about how to present a proposal to a board member who is skeptical of technology investments is partly in the library and partly not. Howard’s documentation of that board member’s concerns is there. His strategic assessment of her institutional position is there. The specific reading of how she processes information, when she is moveable and when she has already decided, what kind of evidence changes her mind and what kind hardens her position: this is present in Howard’s reasoning only partially, because he carries it through attention to a specific person over time, not through a methodology that transfers into structure.

The calibration of when to push a partnership proposal and when to let it sit is not in the library. Howard made several of these judgments during the deployment and the library has records of the outcomes, but not the reasoning process he used to decide when the moment was right. This is tacit knowledge in the classic sense: the practitioner knows how to do it but cannot fully describe how. The AI captured his conclusions. It did not capture the judgment that produced them.

The reading of institutional culture that tells an experienced director when a staff concern is a real operational problem and when it is a normal adjustment response to change: this is not in the library in a form Ellen can use when she needs it. Howard read her staff during the six-month deployment. He said things in session summaries about specific staff members’ concerns that reflected a calibration Ellen did not fully share and the library cannot fully transmit.

Howard and Sonia came back.

The follow-up visit at nine months is now a standard BGO deployment component, built into the structure after the first cohort of deployments showed that the transition from active deployment to post-deployment knowledge library is where the most value is lost. Howard spent three hours with Ellen. He updated two entries in the knowledge library based on the grant application context: the partnership framework section needed a note about how to apply it to a technology grant context that had not come up during the original deployment, and the community needs methodology section needed a clarification about how to weight demographic data when the grant funder’s priorities did not align with the primary community need.

These updates are things the library could not have generated on its own. They required Howard to be present with Ellen’s specific problem and apply his judgment to it in real time. The follow-up visit is not a second deployment. It is a maintenance visit. Ellen called it more valuable than it looked.

Sonia reviewed the community partnership status with the current contact records and found that one of the six active partnerships had shifted into a relationship that needed attention: the partnership coordinator at the city’s workforce development office had changed, and the new coordinator had not been briefed on the partnership framework. Sonia spent ninety minutes with Ellen’s staff on how to restart the relationship with the new coordinator. This also could not have been done from the knowledge library.

What transfers fully, what transfers partially, and what does not transfer at all: the honest synthesis.

The strategic plan transferred. The methodology transferred. The partnership framework transferred in the sense that the relationships it built are active. The knowledge library captures Howard’s reasoning at a resolution that is useful 41 of 43 times. The follow-up visit repairs the specific gaps that accumulated in nine months of use without the deployment active.

What did not transfer is the judgment Howard carries: the specific calibration of this institution’s political dynamics, the reading of individual personalities on the board and among the staff, and the accumulated experience of what kinds of decisions benefit from patience and which ones benefit from urgency. This knowledge is not in the library because it does not exist as a structure. It exists as a practitioner’s embodied understanding of a specific context.

The partial solution to this gap is the ongoing relationship. Howard’s email address. The follow-up visit. The library has both. It does not have the full solution because there is no full solution that fits inside the deployment model’s twelve-week scope and the AI’s knowledge capture capability. The gap is real. The structures built around it are the right partial response.

Ellen submits the grant application. Howard reviews the final draft by email. He catches one place where she has applied the partnership framework to a context it was not designed for and suggests a revision. She revises. The application goes in.

The library could not have caught what Howard caught. His email address is also part of the deployment. The informal persistence of the relationship the structure cannot replace: that is also what the library got.

What Exists Now, What Is Coming, and What Requires Time

#

BGO deployments include the AI knowledge library capability described here. Post-deployment follow-up visits at nine months are now a standard component of the deployment structure. The knowledge library query interface is functional but requires some institutional fluency to use effectively; its limitations are real and the piece names them.

Within one to two years, post-deployment maintenance structure as a standard BGO component: quarterly check-ins, library update protocols, and structured handoff to staff members who were not part of the original deployment and need orientation to what the library contains.

Within three to five years, AI knowledge libraries that update based on institutional use, incorporating the institution’s evolving context into the reasoning structures the library holds. Knowledge inheritance protocols for when the Sage can no longer serve in a maintenance capacity. These are development priorities, not current capabilities.

Howard answers his email. That remains the most important thing the library has.