Main site

Education AI agent · E-Rate

E-Rate Bidding & Certification AI Agent

Keep the E-Rate funding-request file whole — the posted request, the bidding clock, every bid logged against the criteria as published, the discount traced to a dated extract — while a named person certifies.

Typical delivery
4–6 weeks
Deployment
Your stack
Signer decides
Never ranks
After launch
Agent Care

What this agent does

Logs the bids, never ranks them

In

  1. 01

    A request for services is posted, and the waiting period is run from the certification date, not the posting date.

  2. 02

    A material amendment lands mid-window, and it is flagged as one that restarts the waiting period.

Reason

  1. 03

    A bid arrives by email to one engineer, and it enters the same file as every other bid received.

  2. 04

    A bid is read against the evaluation criteria as the applicant published them, and against nothing added since.

  3. 05

    A price is held in its own column, separable, so the primary-factor test can be evidenced rather than asserted.

Decide

  1. 06

    A bid is disqualified, and the file is checked for the reason recorded against it.

  2. 07

    A cost is split between eligible and ineligible components, and the basis for the split is captured with it.

Out

  1. 08

    A CIPA certification falls due, and the policy, the public notice and the minute of the hearing are looked for.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent builds the comparison; the person authorised to order eligible services makes the selection and certifies it under oath.

Example workflow

One funding request, posted to certified

  1. 1Request for services postedPosting record, certification date and the evaluation criteria as published
  2. 2Bids loggedEvery bid received, its response to each published criterion, and its price in a column of its own
  3. 3File assembledBid log, clocks, cost-split basis, CIPA evidence and confidence
  4. 4Controls appliedWaiting-period checks, criteria-drift checks, the procurement rules the district names and confidence threshold

No human action required

Stages 1 to 4 run unaided, and none of them ranks a bid — the agent is logging, and the signer lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the authorised signer to review.

Low confidence

Adds a procurement-counsel read first.

Signer review

The file is held with the criteria as published, the bids as received and the confidence.

Review · Correct · Send to counsel

Reviewed — the signer certifies

  1. 6Procurement record updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedBid-capture completeness, clock accuracy, cost-split basis and findings raised in review

Corrections

Every signer correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Scoring or ranking a bid — that is making the decision the signer certifies they made.
    • Selecting the vendor, or naming the most cost-effective offering.
    • Certifying anything, or signing the funding request.
    • Deciding a cost is eligible where the rules leave it open.

  • Automation boundary

    Agent acts unaided

    • ✓Log every bid received against the criteria as published.
    • ✓Run the waiting period, and flag an amendment that restarts it.
    • ✓Hold price in its own column and check each bid for completeness.
    • ✓Show what the file is missing under the rules the district names, and hold it for the signer.

    Wrong after the oath and the money is already paid; recovery and False Claims Act exposure follow.

    • Determining the discount percentage or the eligibility count behind it.
    • Deciding that a disqualification was properly made.
    • Judging whether state or local procurement law was satisfied.
  • Changes to criteria, clock or certification-control configuration.

Example output

One funding request, annotated

Everything the agent logs is attached to the bid or the criterion it came from.

Bid-log output · single funding request

Illustrative example
Funding request
Internal connections
Logged finding
A bid arrived by email and was logged after the evaluation matrix had been closed
Waiting period
Restart required
Criterion applied
47 CFR 54.503
Confidence
89%
Certification status
Uncertified — bid unlogged

As received

Taken from the posted request and the bids as received — the discount percentage comes from the nutrition office, and our nutrition agent owns that count, not this one.

Record used

  • Posting record
  • Bid as received
  • Published criteria

Why this flag

A material amendment restarts the window, and this bid landed inside the old count — the clock is the finding.

What the score decides

Below the configured threshold the file picks up a counsel read before the signer sees it.

Action

ReviewCorrectSend to counsel

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every funding request

From the posted request

03Logging

Log against the published criteria

Draw on the posted request, the bids as received and the clocks. Wisconsin Bell, Inc. v. United States ex rel. Heath, decided unanimously on 21 February 2025, holds that E-Rate reimbursement requests are claims under the False Claims Act.

01Approved path

The oath is on the selection

Routine bid capture and clock arithmetic arrive already done.

02Human review

Point people at what an auditor will ask for

Unlogged bids, criteria drift and an unexplained disqualification are marked. A 2026 FCC order will make the bid evaluation and vendor selection documents a filing from funding year 2028; a technology plan is not one, and has not been required since funding year 2015.

04

Build an evidence trail

The bid, the criterion it was read against and the person who certified stay on the file.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • E-Rate systems

    • USAC EPC portal
    • Posting and funding records

  • Procurement records

    • Bid inbox and bid log
    • Board agendas and minutes

  • District systems

    • Student information systems
    • Nutrition eligibility extracts

Agent

E-Rate bidding and certification

  • Reads the file
  • Logs the bids
  • Holds for the signer

  • Finance and contracts

    • ERP and purchasing
    • Contract and document stores

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the file

Six graded sieves, coarsest first. What falls through every one is set out in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeHold the funding request at uncertified when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, criteria and clock-configuration changes.Track
L4TraceabilityRecord the bids as received, the criteria applied, the flags and the review time.Record
L3Signer reviewHolds the file for the authorised signer; it governs what is shown, not whether the selection was sound.Gate
L2Policy guardrailsTest each entry against the published criteria and the refusal list; a failure returns the entry.Restrict
L1Confidence thresholdsRoute low-confidence clock and cost-split findings to a counsel read first.Require review
Model coreFile produced — bid log, clocks, cost-split basis and confidence
L1 – L2
Test whether an entry may stand
L3
Puts the file in the hands of the signer
L4 – L5
Keep the bid and the criterion behind it
L6
Holds the request at uncertified when signals degrade

How Nestack evaluates it

Evaluate the whole file — not only the bid log at the end.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the file the signer reads

  1. E1Final-output evaluationWas every bid received in the log, against the criteria as published?
  2. E2Step-level evaluationDid the agent use the right posting, criteria and procurement configuration?
  3. E3Tool evaluationDid it read the correct funding request and write the correct record?
  4. E4Confidence calibrationDo low-confidence entries actually attract more signer corrections?
  5. E5Slice evaluationHow far apart do capture rates sit across specific funding-request types?
  6. E6Business outcomeHow many gaps were corrected, and how many surfaced only in review?

Floor — the certification the district answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, set at the lifecycle stage each one begins in.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • HH-03

      Bid never reached the file

      A bid emailed to one engineer never entered the bid log.

    Stage gathers

    Posting, criteria, bids received and the clocks

  2. 02 · Reasoning

    2 modes

    • HH-04

      Criteria changed late

      A factor is added or re-weighted after the bids were opened.

    • HH-06

      Composite price hidden

      Price is folded into one score and cannot be separated again.

    Stage proposes

    Bid log, clocks, cost-split basis and confidence

  3. 03 · Tool / write

    2 modes

    • HH-02

      Log wired to a ranker

      A completeness flag is piped into a scoring tool elsewhere.

    • HH-05

      Stale discount source

      A live dashboard is read where a dated extract was needed.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • HH-01

      Amendment not detected

      A material change to the posting left the clock counting old days.

    Stage returns

    The file the signer certifies from, and stands behind

  5. 05 · Change / Version

    1 mode

    • HH-07

      Silent criteria drift

      A configuration change alters the criteria and open requests do not recompute.

    Stage tracks

    Model, prompt, criteria rules and clock config

  • Sev-1 · a ranking is produced
  • Sev-2 · a bid is missing at certification
  • Sev-3 · source degrades, file goes to review
See the complete Education failure-mode catalogue

Affected slices

One request type can carry most of the gaps

A district-level bid-capture figure can read sound while a single funding-request type holds most of the unlogged bids and missing bases. Nestack reports the gap rate by funding-request type, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Category two internal connections9.8%3.7×Review
Multi-year contract renewals7.7%2.9×Review
Consortium funding requests4.4%1.7×Watch
Recurring internet access1.6%0.6×Normal

Bar: gap-rate lift vs. the recurring-internet-access baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

A missing bid ends as a test

A cycle closes when the unlogged bid is a regression case. That suite is what the next request assembled is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Gap rate rises in one funding-request type.

  2. 02

    Diagnose

    The vendor who won on a criterion nobody wrote down is read back through the posting and the bid log until the cause narrows to one.

  3. 03

    Improve

    The change ships numbered, and the requests that forced it ride with it.

  4. 04

    Verify

    Nothing releases while one touched bid case is still red.

  5. 05

    Learn

    It is kept for good, and the criterion rules are amended in the same commit.

  6. Learn → Detect

    The return edge. Detection next time meets a suite one case longer than this one.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, bidding workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Bidding workflow discovery and refusal boundary.Week 1
  2. 02EPC and procurement-record assessment.Week 1, Week 2
  3. 03Criteria, clock and cost-allocation rule mapping.Week 1, Week 2
  4. 04Bid and posting ingestion.Week 2
  5. 05Clock arithmetic and criteria binding.Week 2, Week 3
  6. 06Confidence scoring and finding routing.Week 3
  7. 07Authorised-signer review workflow.Week 3, Week 4
  8. 08EPC and procurement-system integration.Week 5
  9. 09Bid and window cases.Week 4, Week 5
  10. 10Guardrails and certification controls.Week 4
  11. 11Bid-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne district, one funding yearProductionProduction EPC and procurement systemsAdvancedMultiple entities / consortia
Introduced at Pilot
Bid logging to your published criteria✓✓✓
Authorised-signer review✓✓✓
Bid-capture baseline✓✓✓
Introduced at Production
Reporting by funding request—✓✓
Review workflow in your systems—✓✓
Approved record write-back—✓✓
Procurement-record integration—✓✓
Introduced at Advanced
Multi-state procurement rules——✓
Multi-stage board approvals——✓
High funding-request volume——✓
Multi-entity certification controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, funding-request volume, review controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your posted requests and where the bids land

    Bid and posting ingestion, and criteria binding

    Week 1

  2. 02Representative files from a past funding year

    Bid-capture baseline, clock normalisation and rule binding

    Week 2

  3. 03Your published criteria and cost-allocation basis

    Criteria, clock and cost-allocation rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    EPC, procurement and eligibility-source review, then integration setup

    Week 2

  5. 05Files you would not want subpoenaed

    Window cases and failure-mode testing

    Week 4

  6. 06What no certification may cover

    Confidence scoring, finding routing, guardrails and certification controls

    Week 3

  7. 07The named person authorised to certify

    Signer review workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each band here is the weeks its own phase truly costs, so the fifth one honestly carries a pair.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Bidding workflow discovery, criteria mapping and the refusal boundary

W2Record integration and the bid-capture baseline

W3Clock arithmetic, cost-split basis and review controls

W4Evaluation suite, window cases and failure-mode testing

W5EPC and procurement integration, a pilot request and targeted corrections

W6One funding year run under the authorised signer, then Agent Care handover

Reading the band
Each bar sits only on the weeks its work is named in. The fifth-week overlap is real, not padding.
At the end of W6
Validation closes on live requests, and Agent Care picks up the watch.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Education AI agent

Build an E-Rate agent around the person who signs.

Show us your posted requests, your published criteria and who holds the authority to certify. If your bid evaluation still lives on one desktop, then know that the FCC calls it a document you are already required to maintain and produce.

Typical delivery
4–6 weeks
Deployment
Your EPC and procurement stack
Signer decides
Never ranks
After launch
Agent Care available

Nestack Agents · E-Rate bidding and certification

AGT-ED-09 · Agent Care available after launch

Build an E-Rate agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your funding requests

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.