Where Your Files Go When You Upload Them to Claude
Uploading a file to Claude is a storage question, not a training one: where it sits, who can open it, how it leaves, and the setup choice that matters.
ReadIndustries · Government
Claude, working on the backlog your agency can't hire its way out of.
Anthropic sells Claude to state and local agencies directly and through its distributor on cooperative vehicles your procurement office already uses. The implementation comes after: Claude in the eligibility file, the records request, and the grant report, with your staff still making every determination. Shannon Advisory does that work with state and local agencies, starting with training your staff on their own work.
The problem
The work your agency can't put off.
Each of these runs on a clock your agency doesn't set: a statute, a funder, a legislature, or a resident waiting for an answer.
- Eligibility caseworkVerifying documents, reading policy that changes at the federal, state, and county level at once, and a queue that refills faster than it clears. Statute sets the volume.
- Public records requestsSearch, review, redaction, and a statutory clock. One broad request can tie up a records officer for weeks.
- Grant applications and reportingChasing the funding a program depends on, then reporting in each funder's format, on deadlines that land on someone with a full-time job.
- Procurement documentsWriting the solicitation, building criteria to match, and scoring what comes back. Months of a program manager's year before a vendor starts.
- Resident and legislative correspondenceInquiries, complaints, and questions in more languages than the office staffs for. Response time is what the public notices.
- Systems nobody left documentation forEligibility on COBOL, permitting on aging .NET, a tax system whose knowledge retired with the person who built it.
Where we fit
Anthropic sells Claude to the public sector. We put it to work inside one of your programs.
In a number of states Claude is already on a contract your procurement office recognizes, including NASPO ValuePoint and OMNIA Partners cooperative contracts, and California has a statewide agreement with Anthropic. Which version you buy follows your data: Claude Enterprise for work that doesn't require FedRAMP authorization, and Claude for Government, Anthropic's FedRAMP High authorized product, where it does. If your state has a vehicle, use it. Access comes through Anthropic, not us. We don't sell, broker, or discount it.
The contract gets you seats, a training offer, and a capable tool on every desk. Take the training. What's left is fitting the eligibility-review workflow to your policy manual, citation format and sign-off step. In government it also has to hold up to a records request, an audit, and staff turnover.
The caseworker, the records officer, or the program lead makes every determination that affects a resident, and the workflow enforces that. Every output carries the policy section it came from, so a caseworker can check it before it goes out and an auditor can see what it rested on.
Workflows
Six workflows inside a public agency.
All six are set up on your own policy, templates, and prior responses, and each ends with a named staff member signing off.
- Eligibility and benefits caseworkPolicy questions answered against the federal, state, and county rules that govern the case, with the citation attached. The caseworker still makes the determination.
- Public records requestsSearch across the record set, a first pass for responsiveness, and proposed redactions with the exemption cited for each. The records officer reviews and signs off.
- Grant applications and funder reportingApplications drafted against the funder's criteria from documents you already have, and the recurring report in each funder's format. The program lead still owns the narrative and the numbers.
- Procurement documentsA draft solicitation in your jurisdiction's template, with evaluation criteria that follow from the requirements. Scoring stays with the committee. If drafting it would conflict with our role later, your procurement office draws the line.
- Resident and legislative correspondenceInquiries, complaints, and resident questions drafted from your own policy and prior responses, in plain language and in the languages residents write in. Staff review and send.
- Systems nobody left documentation forReading an undocumented system and writing down how it works, writing the tests nobody wrote, and scoping a change so you can put a risk estimate in front of the budget owner.
Proof
Why us
Claude is the only AI we build with or train teams on.
The people on our client engagements hold Anthropic's Claude certifications. Every case study we've published is on the Work page.
Shannon Advisory is a Select partner in the Claude Partner Network Services Track.
How we work
One program first, on the Claude contract you already have.
We start with one workflow for the queue that's furthest behind, set up on your policy and tested on real cases before it goes live.
What we do
Adopt
Training on your own work, setup with your IT team, connecting the systems Claude connects to, weekly office hours, and skills written with your team.
See AdoptBuild
Custom connectors for systems Claude doesn't connect to yet, and apps that replace a manual routine.
See BuildA first engagement runs about three months. Most clients keep us on after that, by the quarter or the year. Pricing by proposal.
FAQ
What agencies ask before the first call.
We're a state agency, not a federal one. Does this apply to us?
Yes. Shannon Advisory works with state agencies, counties and city departments. Our work is a separate, much smaller engagement than the Claude contract, on whatever instrument your office prefers: a professional services roster, a small purchase, or whatever your procurement rules allow.
How does a state or local agency buy Claude?
Directly from Anthropic, or through its distributor on cooperative contracts such as NASPO ValuePoint and OMNIA Partners. California state entities, cities, and counties can buy through the California Department of Technology's Statewide IT Shared Services program, under agreement STP-GENAI-26-ANTHROPIC. The purchase is between your agency and Anthropic, and Anthropic's public sector page is the place to start. We don't sell, broker, or discount it.
What about security and compliance?
Anthropic runs the compliance program behind Claude, and the authorizations attach to the environment Claude is deployed in. Claude Enterprise is for government work that doesn't require FedRAMP authorization. It isn't FedRAMP authorized today, and Anthropic lists FedRAMP Moderate for it as in process. Claude for Government, Anthropic's FedRAMP High authorized product, is available to state and local agencies, as is Claude through Amazon Bedrock on GovCloud or Google Vertex with Assured Workloads. Agencies handling criminal justice information (CJIS) or federal tax information (IRS Publication 1075) typically use one of those. A HIPAA-ready configuration with a business associate agreement is available for programs handling protected health information. We hold none of those authorizations. Before anyone writes a prompt, we map what stays in the agency, what goes to Claude, and which posture your program's obligations require.
How do Claude prompts and outputs fit our records retention?
Your records officer decides how your records rules apply. We set up the workflow so the prompt and its source citation stay attached to the work product, retention follows your existing schedule, and a records officer can produce the output months later. Claude for Government keeps conversation history on the agency-managed device, so on that product the workflow saves the prompt and output to your records system.
Where does Claude actually help, and where should it stay out?
The reading and the assembly, not the determination. Anthropic's usage policy also prohibits several uses that come up in government, including determinations about criminal justice matters such as parole or sentencing eligibility, predictive policing, and tracking a person's location, emotional state, or communications without their consent, which is where facial recognition and biometric categorization sit. Anything that goes to a resident is reviewed and sent by a qualified staff member, and the file records that AI helped produce it. A resident working with Claude directly is told so at the start.
Our procurement cycle is long. How does that affect the timeline?
Procurement usually takes longer than the implementation. If your state or a cooperative vehicle already carries Claude, that part is much shorter. Our engagement is separate from the Claude purchase, so we can scope it while procurement runs. A first engagement runs about three months, and most clients keep us on after that, by the quarter or the year. The proposal sets the binding timeline.
Bring the queue that isn't moving. We'll start there.
Or bring the records request you know is coming. We'll tell you if Claude is the wrong tool for it.