AI Tools for Corporate Pension Research
Choose AI tools for corporate pension analysis. Use IBM's disclosures to distinguish funded status, plan deficits and cash contributions in an investment review.
In this article
AllMind fits corporate pension research when an equity or credit team needs to connect plan disclosures with its investment assumptions across a coverage list. Our Grids applies reusable questions to companies and retains source-linked answers. Data Rooms keeps the selected filings alongside the firm's earlier research. The output should explain which pension obligations matter to the issuer, what changed and what the evidence actually says about funding.
For an analyst starting with standardized financial statements, Capital IQ Pro is another candidate. For a firm assembling a specific collection of annual reports, actuarial material and internal analysis, evaluate Hebbia. None of these choices removes the need to distinguish a plan's accounting position from a company's near-term cash requirement.
IBM provides a useful test. Its year-end 2025 defined-benefit pension positions add to a $474 million net surplus, while its individual underfunded pension categories have $7.071 billion of deficits. Both descriptions can be correct. A separate 2026 contribution table introduces another distinction: most of the reported amount was contributed in securities, rather than cash.
Disclosure: We develop AllMind and benefit if a team chooses it. This is a documented comparison, last checked November 16, 2026, with our calculations from public filings. We did not run the same pension task in each product. Competitor capabilities are vendor-reported; the example establishes requirements for the research, not measured product performance.
Choose the tool around the decision
Corporate pension analysis has several possible starting points. A cross-company screen can identify names with large retirement obligations. An issuer review must then explain the plans, the funding arrangements and the effect on the house view. An actuarial valuation is a further, specialist task.
| Research requirement | Candidate | Source-supported capability and buying boundary |
|---|---|---|
| Revisit pension questions across issuer coverage and relate the results to the firm's credit or equity thesis | AllMind | Cited company questions in Grids and selected filings plus internal research in Data Rooms. Confirm the relevant disclosures and retain the plan-level exceptions. We do not establish an actuarial valuation capability here. |
| Start with standardized financials and investigate the source disclosures in the same financial-data environment | Capital IQ Pro | Its official product page describes financial data, ChatIQ and Document Intelligence. Confirm that the required pension fields, dates and source documents are covered. |
| Analyze an assembled document collection using the firm's own research process | Hebbia | Its product overview describes Matrix analysis across documents and companies with traceability. Its platform supports firm-document workflows and Office plug-ins. Test the handling of plan groupings and table footnotes on the actual packet. |
These are starting points, not exclusive feature categories. All three can be relevant to document-based work. The deciding issue is whether the configured workflow preserves the distinctions below and delivers the output the investment team needs. We have not verified a common subscription package, complete pension-field coverage or comparative accuracy.
For a single issuer, the filing and a checked spreadsheet may finish the assignment. A wider research platform earns its place when the same questions recur across companies and must connect to prior investment work.
A net surplus can coexist with substantial deficits
IBM's 2025 retirement disclosure separates defined-benefit pension plans by geography, qualification and funding position. The following figures are at December 31, 2025, in millions of dollars.
| Defined-benefit pension category | Reported surplus / (deficit), $ millions |
|---|---|
| U.S. qualified pension plan | 4,866 |
| U.S. nonqualified pension plans | (1,253) |
| Overfunded non-U.S. qualified plans | 2,679 |
| Underfunded non-U.S. qualified plans | (1,274) |
| Non-U.S. nonqualified plans | (4,544) |
| Net across these pension categories, our sum | 474 |
The underfunded categories total $7.071 billion, calculated as $1.253 billion plus $1.274 billion plus $4.544 billion. The surplus categories offset those deficits in the arithmetic. The calculation uses rounded reported category amounts; it excludes nonpension postretirement benefits.
That distinction changes an investor's next question. “The company has a pension surplus” is incomplete if it causes the analyst to stop reviewing the underfunded plans. “The company must immediately pay $7.071 billion” is also unsupported. An accounting deficit does not by itself specify the timing or size of legally required contributions.
The source describes contributions to the U.S. qualified plan as held in an irrevocable trust for participants and beneficiaries. A surplus there should therefore not be treated as an unrestricted corporate cash balance. The table also does not establish that assets can be moved between all these plans to extinguish other obligations.
For a valuation or credit note, those are separate issues: the net accounting exposure, where deficits remain and the evidence for future funding. An AI summary that collapses them into one number can obscure the question the investor was trying to answer.
The contribution table is not a cash-flow statement
The latest quarter adds a second test. IBM's September 2026 Form 10-Q, Note 18 reports these contributions for the nine months ended September 30, 2026:
| Items in the disclosed contribution table | $ millions |
|---|---|
| U.S. nonpension postretirement benefit plan | 187 |
| Non-U.S. defined-benefit and multi-employer plans, net of refunds | 22 |
| Total shown | 209 |
The footnote says the $187 million U.S. nonpension contribution was made in U.S. Treasury securities, which IBM treats as a non-cash transaction. Subtracting it from the table's $209 million total leaves $22 million in the other category. Calling the entire $209 million a cash pension payment would therefore misstate both the form and the benefit category.
The table is not a complete measure of all retirement-related cash spending. It does not combine every defined-contribution payment, direct benefit payment or other benefit arrangement. The same footnote separately identifies Treasury-security contributions to an Active Medical Trust. Adding selected rows without understanding their scope would create a new problem.
This matters when a research model starts with reported operating cash flow and proposes a pension adjustment. An analyst needs to know whether the adjustment is a cash item, a non-cash transfer, an expense or an accounting liability. Subtracting a non-cash securities contribution as an additional cash payment would not be justified by this table.
The quarter also does not remeasure the annual table above for us. The pension funding positions retain their December 2025 date; the contribution figures cover nine months of 2026. Combining them in one undated “current pension liability” column would imply a precision the documents do not provide.
Funding expectations answer a different question
IBM's September filing says it was not legally required to contribute to its U.S. defined-benefit plans during 2026, while noting that market performance could increase minimum contributions in some non-U.S. countries. This is management's reported funding position for that year, not a finding that pension obligations disappear or that future contributions will always be immaterial.
For an investment team, the useful distinction is between an obligation and its financing timetable. A balance-sheet deficit can remain relevant to long-term value even when the current year's required contribution is small. Conversely, assuming a near-term cash call equal to the entire deficit can distort a liquidity review.
The same separation applies to pension expense. Service cost, interest cost, expected asset returns and actuarial items can enter different parts of the financial statements. A cash-flow forecast should not silently substitute one of those expense measures for actual contributions.
We have not calculated IBM's statutory funding requirements or a legal transfer of assets between plans. That would require the relevant plan rules and jurisdiction-specific work. The supported conclusion is narrower: the disclosures contain different measures, and the research must retain their definitions before deciding how they affect the issuer.
Where AllMind belongs in the coverage process
For this job, the most useful AllMind Grid is a company comparison that keeps the plan category, funded-status date, reported deficit, contribution period and funding form separate. A column asking only for “pension liability” would lose much of the information demonstrated above.
The firm's prior note supplies the investment question. Perhaps its liquidity case assumes that pension funding stays modest; perhaps its valuation adjustment nets every pension asset against every obligation. A Data Room containing the selected company disclosures and that note gives the follow-up a defined scope: which assumptions are supported, which need qualification and which require further evidence?
Our Reports can turn the selected evidence into an editable, cited section in the team's investment review. The finished artifact should preserve the pension-only calculation, the current contribution disclosure and the unresolved funding questions. That combination is the reason we recommend AllMind for repeat issuer research, rather than for a standalone actuarial calculation.
Keep a specialist model and adviser where the decision requires statutory minimum funding, benefit valuation or plan-specific legal interpretation. For a broad company-data screen, an existing financial-data platform may be the natural first step. For a firm-owned document workflow, Hebbia is a credible alternative to evaluate. The choice should follow the work being repeated.
The IBM packet makes a concrete acceptance test: preserve the $474 million net pension position alongside the $7.071 billion underfunded categories, identify the non-cash portion of the contribution table, and avoid relabeling either as a current cash requirement. Then assess whether the output improves the firm's actual investment argument.
Related work continues in our debt-refinancing research guide, where the reporting entity and maturity date change the financing question, and our investment-thesis review guide, where new evidence is tested against a prior assumption.
Sources and scope
The worked example uses IBM's 2025 retirement disclosure and September 2026 Form 10-Q, accessed November 16, 2026. Figures retain the benefit categories, dates and units in those filings. The net position, gross deficits and contribution subtraction are our calculations. They are not an assessment of IBM's creditworthiness or a forecast of future pension funding.
The linked official product pages support the documented shortlist. No product was tested on this packet for the article, and no pension-field coverage or calculation performance is assumed from general document-analysis capabilities.