Track Management Commitments and Follow-Up Evidence Across Earnings Calls With Claude MCP

A management commitment register connects an operating target to the evidence reported after it. It preserves what management originally promised, records the deadline, and tracks whether subsequent disclosures measure the same outcome. That makes it easier to distinguish a completed commitment from a larger aggregate total, a revised plan, or a statement that leaves the original quantity unconfirmed.

Management teams often revisit the same initiatives across several earnings calls. A plan to close specific facilities becomes a report on total closures. A volume-reduction target becomes a statement that a broader program is complete. Those updates may show substantial progress, but their relationship to the original commitment needs to be established.

This workflow uses Claude and Financial Modeling Prep's MCP server to build a register for United Parcel Service (UPS) across three consecutive earnings calls. Claude extracts numeric operating commitments from the first call, then compares three selected commitments with later commentary and the second-quarter 10-Q.

The resulting register distinguishes three situations: management reports completion without a directly comparable quantity; a filing confirms closures above the original count without identifying the original buildings; and a transcript figure needs clarification. Each entry records what the available evidence supports and the question a subsequent disclosure would need to answer.

The scope is operating commitments. Verifying management claims against reported financials addresses a related task involving revenue, margins, cash flow, and other financial measures.

Key Takeaways

• Preserve the original objective and deadline alongside later updates. A new statement should not overwrite the terms being evaluated.
• Compare the same unit, population, and measurement period. A broader total can exceed a target without proving that a specific group of items was completed.
• Keep management's completion statements separate from directly reported quantities and calculated comparisons.
• Distinguish clear evidence with limited scope from evidence whose meaning is uncertain.
• An unresolved entry is a valid register result when it explains what is known, what remains unconfirmed, and what evidence would resolve the question.

Define the Commitments and Evidence Rules

UPS provides a useful example because management discussed the same network initiatives across successive calls, while its second-quarter filing supplied an additional disclosure on building closures.

The example covers fiscal Q4 2025, Q1 2026, and Q2 2026. The retrieval took place on October 22, 2026. Call dates, fiscal periods, and the periods measured by individual statements remain separate throughout the register. Keeping those distinctions explicit also helps when aligning earnings dates, fiscal quarters, estimates, and reported results.

Select Numeric Operating Commitments

This register focuses on operating statements with a numeric objective and a stated time frame. That is a selection rule for the example, rather than a claim that every measurable commitment must contain a number.

A dated facility opening, for instance, can be a concrete milestone even without a quantity. It falls outside this example's numeric-target scope.

Statement category	Treatment in this workflow
Numeric operating objective with a stated time frame	Candidate for the register
Concrete nonnumeric milestone	Outside this example's scope
Broad ambition without a defined outcome or time frame	Record separately as an aspiration
Revenue, margin, earnings, savings, cash-flow, or capital-spending target	Exclude as financial guidance
Unclear objective or timing	Flag the ambiguity before selection

The extraction can produce several statements about one initiative. Aircraft deliveries across different periods, for example, should remain linked rather than being counted as unrelated promises.

Preserve Original Terms and Later Updates

Each selected commitment needs an original objective, its timing language, and enough surrounding context to interpret both. Later changes belong in separate fields.

For the UPS example, the Amazon volume target refers to full-year 2026, while surrounding remarks describe the final six months of the accelerated program. Both references remain visible. Subsequent statements about June completion are recorded alongside them.

A later deadline should be identified as a delay relative to the original commitment. An earlier date should also be preserved, even when it falls within the original window. Neither update should erase the starting terms.

Match the Measure Before Assigning a Status

A direct comparison requires the same unit, population, and measurement period. Two additional checks matter in this example.

Identified groups of items. If management says it has identified 24 buildings for closure, a later total of 45 closures establishes that the aggregate count exceeds 24. It does not, by itself, identify whether the original 24 are included.

Calculated comparisons. Subtracting one reported figure from another requires compatible definitions. The calculation should preserve each input's source, population, starting point, and measurement basis. Correct arithmetic cannot resolve an uncertain comparison.

The register also distinguishes completed actions from levels measured at a particular date.

Commitment type	Timing treatment
One-time action, such as closing buildings	Can be classified as achieved before the deadline if matching evidence establishes completion
Level evaluated at a target date, such as year-end automation share	Requires evidence describing that target date or period

For automation, this workflow conservatively evaluates the year-end level. A confirmed earlier reading can show progress, including a level above the target, without establishing the year-end result. A disclosure published after year-end is relevant only if it actually describes the required measurement date or period.

Separate Evidence Quality From Evidence Coverage

A statement can be clear and credible while answering only part of the question. That is different from a garbled figure or a conflict between sources.

Evidence type	Meaning
Management-reported	The measure or completion statement appears in call commentary
Filing-corroborated	A filing separately reports the relevant measure
Filing reference without measure	A filing discusses the initiative without reporting the quantity needed

A filing remains a company disclosure. “Filing-corroborated” means the measure appears in a separate regulated document, not that an independent party has verified every underlying event.

The following statuses describe the evidence available for a commitment.

Status	When to use it
Review Required	A material figure is unclear, comparable evidence conflicts, or the meaning of the evidence cannot be established
Achieved	Directly comparable evidence establishes the objective under the applicable timing rule
Exceeded in aggregate, inclusion not established	A broader count exceeds the target, but completion of the original identified group is unconfirmed
Completion reported, objective not directly evidenced	Management reports completion without directly reporting the commitment's own quantity
Delayed	Evidence establishes that the original deadline was missed or moved later
Progress reported	Comparable progress is reported, but completion under the applicable timing rule is not established
Not Updated	No relevant follow-up evidence was found

Clear aggregate evidence does not automatically trigger Review Required. Neither does an explicit completion statement that lacks the original quantity. Those situations have their own classifications.

A calculation whose basis is unconfirmed remains separate from the status evidence. Silence alone does not establish delay or failure.

Connect Claude to the FMP Data Sources

The workflow uses FMP's earnings call transcripts and filing disclosures. The connector operation names below are the names used in the example; the documentation identifies their corresponding API routes.

Connector operation	Inputs	Purpose
earningsTranscript / transcripts-dates-by-symbol	symbol=UPS	List fiscal periods and call dates
earningsTranscript / search-transcripts	symbol=UPS; year; quarter	Retrieve each call's text
statements / financial-reports-form-10-k-json	symbol=UPS; year=2026; period=Q2	Retrieve the quarterly filing content

Transcripts Dates By Symbol supplies the company-specific call dates. The Earnings Transcript API retrieves the text for the selected year and quarter.

The filing operation's name includes “10-K,” but the example requests Q2 and checks the returned document type. FMP's financial-report JSON documentation describes the underlying financial-reports-json route, and its financial-report listings include quarterly requests.

Use an FMP plan with access to the required transcripts and filing data. Following the FMP MCP connection instructions, open Claude's Settings, select Connectors, and add a custom connector using this URL with your API key:

https://financialmodelingprep.com/mcp?apikey=YOUR_FMP_API_KEY

Start a new conversation with web search off. Run the first prompt, select the commitments, and then run the second prompt in the same conversation so Claude can reuse the original transcript.

Extract the Original Commitments

The first prompt separates extraction from follow-up evaluation. Making the selection before retrieving later evidence helps reduce the risk of choosing only commitments with convenient outcomes.

Prompt 1: Identify Candidate Commitments

Using Financial Modeling Prep through the connected FMP MCP server, start a management commitment register for United Parcel Service (UPS). Objective Identify measurable operating commitments made on the UPS fourth quarter 2025 earnings call and separate them from broad aspirations and financial guidance. Do not track follow-ups yet. Data rules - Use only the FMP earningsTranscript tool. Do not use web search. - Quote statements verbatim, 20 words or fewer. Keep transcript errors as they appear and flag any bracketed or unclear figure. - If a request fails, retry it once. If a transcript is still missing, report Unavailable. Steps 1. Call transcripts-dates-by-symbol for UPS. Confirm that the calls for fiscal Q4 2025, Q1 2026 and Q2 2026 exist and are consecutive, list their call dates, and report the most recent call listed. 2. Call search-transcripts for UPS, year 2025, quarter 4. 3. List forward-looking statements about operations or targets and classify each one: - Measurable commitment: a numeric operating objective with a stated deadline - Broad aspiration: a direction or ambition with no measure or no deadline - Financial guidance: revenue, margin, earnings, savings, cash-flow or capital-spending targets (out of scope) 4. For each statement, record the speaker, the objective and the deadline exactly as stated. Do not turn a relative deadline into a date unless the transcript does. Output Start with a short run summary: run date, company, calls in scope, operations called and any data limitation. Table 1: Fiscal period | Call date | Transcript available | In scope Table 2: # | Statement (verbatim) | Speaker | Measurable objective | Deadline as stated | Classification End with one line listing the measurable commitments. Do not select or evaluate them.

Calls and Targets in the Example

The October 22 retrieval listed the following calls. Prompt 1 retrieved only the first transcript; Prompt 2 retrieved the later two.

Fiscal period	Call date	Role
Q4 2025	January 27, 2026	Original commitments and baseline figures
Q1 2026	April 28, 2026	First follow-up
Q2 2026	July 28, 2026	Second follow-up

The extraction surfaced eight qualifying numeric statements, including overlapping aircraft-delivery schedules. They covered Amazon volume reduction, automated volume share, additional automated buildings, operational hours, building closures, and aircraft deliveries.

Three commitments form the worked register:

	Operating objective	Original timing
Amazon Volume Reduction	Reduce Amazon volume by another 1 million pieces per day	“for the full year 2026”
Building Closures	Close 24 buildings identified for closure	“in the first half of the year”
Automated Volume Share	Increase U.S. volume processed through automated facilities to 68%	“by the end of the year”

The two building-related targets in the extraction should not be confused. Management discussed both adding 24 automated buildings and closing 24 buildings. The transcript did not establish that these were the same facilities.

The selected commitments cover three distinct comparisons: a quantity reduction, completion of an identified group, and a level measured at a target date.

Match Later Evidence to Each Commitment

The second prompt retrieves the follow-up calls and the quarterly filing. It preserves the original terms while separating directly comparable evidence, broader totals, completion statements, and calculations.

Prompt 2: Build the Evidence Register

Continue the UPS commitment register. Track these three commitments from the fourth quarter 2025 call, identified by their exact wording: C1 (Amazon glide-down): "for the full year 2026, we intend to glide down another million pieces per day" C2 (building closures): "we have identified 24 buildings for closure in the first half of the year" C3 (automation share): "increase the percentage of U.S. volume we process through automated facilities to 68% by the end of the year" Objective Match later evidence to each commitment, keep its original terms separate from later restatements, and assign a status under the rules below. Data rules - Use only the FMP earningsTranscript tool and the FMP statements tool. Do not use web search. - Transcripts: UPS fiscal Q1 2026 and Q2 2026 for follow-up evidence, and fiscal Q4 2025 only for baseline figures stated on that call. - Filing: call the statements tool with endpoint financial-reports-form-10-k-json, symbol UPS, year 2026, period Q2, and confirm the document type is 10-Q. - Quote evidence verbatim, 20 words or fewer. Flag bracketed or unclear figures. - Retry a failed request once; report Unavailable if still missing. Steps 1. Commitment type: classify each commitment as a one-time action (a count or quantity to complete) or a level target (a level that must stand at the deadline). 2. Deadlines: record the original deadline exactly as stated, including every time reference in the original statement or its surrounding remarks on the same call. Record any later restatement or change separately, with its source and date. A restatement that keeps the same date or names an earlier one is a restatement, not a move. The status uses the original deadline; if management later moved it to a later date, record the move. 3. Matching: a statement is a direct match only if it measures the same unit, population and period as the commitment. If the commitment names a specific set of items, such as buildings already identified, a direct match must show that set was completed; a larger total that does not identify those items is an aggregate match. Where evidence separates permanent from temporary completion, report both figures. A statement that a program or action is complete, without the commitment's own figure, is not a direct match for a quantity or count; record it as a completion statement. Log statements on the same theme that measure something else as Not matched, with the reason. 4. Derived figures: if the commitment's figure is not reported directly but can be calculated from reported figures, show the arithmetic and each input's source and date, and test whether the inputs share unit, population, starting point and measurement basis. Record the result as Consistent basis or Basis not confirmed, with the reason. Record any management statement that the action is complete as separate evidence. 5. Evidence type: Management-reported (call), Filing-corroborated (the 10-Q states the same measure), or Filing reference without measure. 6. Status, applying the first rule that fits: a. Review Required: evidence conflicts, a figure the status depends on is bracketed or unclear in the transcript, or a direct match is uncertain. Basis not confirmed for a derived figure does not by itself make a commitment Review Required. b. Achieved, filing-corroborated or Achieved, management-reported: a direct match shows the objective met. A one-time action can be Achieved before its deadline once completion is shown. A level target can be Achieved only on evidence dated at or after its deadline. c. Exceeded in aggregate, inclusion not established: an aggregate match exceeds the target, but no evidence shows the named items were completed. State the evidence type. d. Completion reported, objective not directly evidenced: management states the action is complete, but the objective's figure is not reported directly and any derivation has Basis not confirmed. e. Delayed: the deadline passed and the evidence shows the objective was not met by it, or management moved the deadline later. f. Progress reported: the deadline has not been reached and a progress figure is reported. Make no claim that the deadline will be met. g. Not Updated: no matched evidence found. Silence is not failure. 7. Write one next review question per commitment. Output Table 3 columns: Commitment | Source | Date | Evidence (verbatim) | Match | Evidence type Table 4 columns: Commitment | Commitment type | Original objective | Original deadline | Later restatements (source) | Deadline used for status Table 5 columns: Commitment | Matched evidence | Derived comparison | Evidence type | Status | Next review question After the tables, explain each status other than Achieved in at most two sentences. Do not assess management credibility or guidance.

The example retrieved both follow-up transcripts and the filing on the first attempt. The filing response identified a 10-Q for the period ended June 30, 2026. Relevant disclosures appeared in the Property, Plant and Equipment and Transformation Strategy Costs notes.

The sections below regroup Claude's evidence log, deadline table and register by commitment; the quotations, figures and statuses are as returned.

The response did not supply a filing publication date or URL. The corresponding UPS second-quarter 2026 filing provides a public reference for the disclosures discussed below. This example is a retrospective register, not a reconstruction of exactly what was available on each call date. That separate task requires checking filing availability as of a specified date.

Amazon Volume Reduction

The original objective was another 1 million pieces per day of Amazon volume reduction for full-year 2026. On the same call, management described the final six months of its accelerated glide-down plan. Those contextual remarks matter when interpreting subsequent references to June.

Original Terms and Later Updates

Field	Recorded terms
Objective	Reduce Amazon volume by another 1 million pieces per day
Commitment type	One-time quantity reduction
Original timing	Full-year 2026, with same-call references to the final six months of the accelerated program
Q1 update	Management said the program would wrap up by the end of June
Q2 update	Management reported completion in June

The June references fall within the original full-year window. They remain visible alongside the original wording rather than replacing it.

Evidence Across the Calls

Source	Reported evidence	What it establishes
Q4 2025 call, January 27, 2026; Carol Tomé	Approximately 1 million pieces per day reduced by the end of 2025	Prior-year baseline
Q1 2026 call, April 28, 2026; Carol Tomé	An average reduction of 500,000 pieces per day during Q1	Partial-period progress
Q2 2026 call, July 28, 2026; Carol Tomé	Approximately 2 million pieces per day eliminated over a broader period	Cumulative quantity, not a directly reported 2026-only result
Q2 2026 call, July 28, 2026; Brian Dykes	The Amazon glide-down and related network reconfiguration were completed as planned; completion was in June	Explicit management completion statements
Q2 2026 10-Q, Transformation Strategy Costs note	Discussion of planned volume declines from the largest customer	Filing reference without an Amazon pieces-per-day measure

The Q1 transcript describes the volume as “nonnutritive Amazon volume.” That wording is retained as a transcript limitation; the register does not silently replace it.

Several related figures do not answer the original question. Amazon's 8.8% share of total revenue is a revenue measure. The 3.3% decline in total U.S. average daily volume covers a broader population. Neither establishes the Amazon-specific reduction.

Why the Subtraction Does Not Confirm the Quantity

The reported figures suggest this arithmetic:

Approximately 2.0 million pieces per day cumulative reduction − approximately 1.0 million pieces per day reduced in 2025 = approximately 1.0 million pieces per day.

The numerical result resembles the 2026 objective, but the comparison's basis is unconfirmed.

Comparability check	Unresolved point
Population	References to Amazon volume, lower-quality Amazon volume, and lower-yielding volume are not established as identical
Starting point	Tomé refers to the announcement 18 months earlier; Dykes refers to the beginning of 2024
Measurement basis	The approximate inputs do not establish a consistent exit-rate or average basis

The calculation is therefore illustrative. It cannot independently confirm the 2026-only quantity.

Status: Completion reported, objective not directly evidenced.

Management explicitly reported completion, but the reviewed sources did not directly report a comparable 2026-only reduction in pieces per day. The filing discusses the initiative without supplying that measure.

Next review question: Does UPS disclose the 2026-only Amazon reduction in pieces per day, with a definition and starting point that can be compared with the original objective?

Because the initiative concerns a major customer, its broader implications can also be examined by comparing disclosed customer concentration across annual filings. That analysis should remain separate from whether this particular volume commitment was fulfilled.

Building Closures

The original commitment identified 24 buildings for closure in the first half of 2026. Subsequent disclosures reported more closures than that count, including confirmation in the 10-Q.

Original Terms and Later Updates

Field	Recorded terms
Objective	Close the 24 buildings identified for closure
Commitment type	One-time action involving an identified group
Original deadline	First half of 2026
Q1 update	Plans for 27 additional closures during the year, most in Q2
Q2 update	Further closures planned for the second half

Additional planned closures expand the program. They do not identify which facilities were included in the original group.

Reported Closure Counts

Source	Measurement period	Reported count
Q1 2026 call, April 28, 2026; Brian Dykes	First quarter of 2026	23 buildings closed
Q2 2026 call, July 28, 2026; Brian Dykes	First half of 2026	45 buildings closed
Q2 2026 10-Q, Property, Plant and Equipment note	Six months ended June 30, 2026	45 buildings closed, including 44 permanently

The first-quarter and first-half figures are compatible. The second-quarter count was not separately reported in the cited call evidence.

Both the total closure count and the permanent count exceed 24. However, the original statement did not explicitly require permanent closure, so the 44 permanent closures provide additional detail rather than a new condition for evaluating the promise.

The remaining question concerns the group of buildings. Neither the reviewed call statements nor the filing identifies whether the original 24 are among the 45 reported closures.

Status: Exceeded in aggregate, inclusion not established.

The filing corroborates a first-half closure count above the original target. It does not establish completion of the original identified group.

Next review question: Does a subsequent disclosure establish that the 24 buildings originally identified are included in the 45 first-half closures?

This distinction preserves both findings: UPS reported more closures than the original count, and the evidence reviewed does not identify the original facilities within that total.

Automated Volume Share

The automation objective was to increase the percentage of U.S. volume processed through automated facilities to 68% by year-end, from 66.5% at the end of 2025.

For this register, achievement is evaluated against the year-end level. Earlier readings remain relevant progress evidence.

Original Terms and Follow-Up Evidence

Source	Evidence	Interpretation
Q4 2025 call, January 27, 2026; Carol Tomé	68% objective; 66.5% at the end of 2025	Target and baseline
Q1 2026 call, April 28, 2026; Carol Tomé	“We now have automated 67.5 points our teams almost on our way to 68%.”	Wording is unclear
Q2 2026 call, July 28, 2026; Carol Tomé	“[ 68.5 ]” for volume flowing through an automated building by the end of Q2	Bracketed figure with no percent sign in the quoted passage
Q2 2026 10-Q, Transformation Strategy Costs note	Discussion of network automation	No automated-volume-share measure

The second-quarter transcript also compares the result with 64% a year earlier. That refers to the end of Q2 2025, whereas the 66.5% baseline refers to the end of 2025. The two numbers describe different dates and do not conflict on that basis.

The bracketed 68.5 figure should not be silently converted into a confirmed percentage. Nor does a statement about additional packages passing through automation independently establish the share of total U.S. volume.

Status: Review Required.

The available transcript wording does not support treating the reported Q2 level as an unambiguous percentage. Even if that reading were confirmed, it would establish an interim level rather than the year-end result used by this register.

Next review question: Can the Q2 figure be confirmed as a percentage of U.S. volume, and what comparable level does UPS disclose for year-end 2026?

Read the Completed Register

The three entries produce different conclusions because their evidence has different strengths and limitations.

Commitment	Status	Evidence boundary
Amazon volume reduction	Completion reported, objective not directly evidenced	Management reports completion, but the 2026-only quantity is not directly established
Building closures	Exceeded in aggregate, inclusion not established	The filing confirms 45 closures, but inclusion of the original 24 is unconfirmed
Automated volume share	Review Required	The interim transcript figure needs clarification, and year-end measurement remains outstanding

None is classified as Achieved under this register's rules. That does not establish that UPS failed the commitments. It identifies the conclusions that can be drawn from the selected sources without filling gaps through inference.

The building entry has clear numerical evidence whose coverage is limited. The Amazon entry has clear completion statements without a directly comparable quantity. The automation entry has uncertainty in the reported figure itself. Keeping these situations separate makes the register more informative than a single pass-or-fail score.

Review Required applies to the affected entry; it does not invalidate the completed register. Its practical value is to identify the source question that must be resolved before making a stronger claim.

Update the Register as New Evidence Arrives

Each subsequent call or filing can add evidence without changing the original commitment. Preserve the source, publication or call date, measurement period, quotation, interpretation, and status assigned at that stage. Retaining earlier versions supports reproducibility in financial research, particularly when transcript records are corrected or disclosures change.

For this example, the next quarterly disclosures may clarify the Amazon quantity or building population. Automation requires both clarification of the interim figure and evidence describing the year-end level. If a cleaner source resolves a transcript ambiguity, record the new source and the reason for the status change instead of silently replacing the earlier excerpt.

Ownership can change as well. Where a commitment extends beyond a leadership transition, recording executive-transition context helps preserve who made the original statement and who later reported on it. A change in leadership does not itself establish whether the commitment was completed.

The register gives each new disclosure a specific question to answer. A larger number, a completion statement, and a clarified measurement can each advance the analysis, but only when their relationship to the original objective remains visible.