CASRAI Dictionary

Tag: research administration software

  • AI in Research Administration: Where It’s Actually Deployed

    Most coverage of artificial intelligence in higher education still centres on the classroom — chatbots writing essays, detectors chasing them. Less visible, but arguably more consequential for research offices, is AI in research administration: the back-office layer of proposal budgeting, compliance screening and post-award reporting that keeps federally and privately funded research compliant and auditable. That layer is where AI is quietly moving from pilot to production in 2026, and the evidence — not the marketing copy — shows a narrower, more cautious footprint than headlines suggest.

    This is not a piece about generative AI and authorship integrity, disclosure norms, or research misconduct detection in manuscripts — those questions sit in a separate, already well-documented debate. This is about the administrative machinery: proposal-budget checking, risk-based compliance review, contract redlining and financial reporting inside research offices, sponsored-programmes units and grants-management systems.

    Where AI Is Actually Being Deployed

    The clearest signal comes from a March 2026 Ithaka S+R report, funded through the National Science Foundation’s GRANTED programme (grant #2437518), which convened two workshops — one at Montclair State University (31 participants, 13 institutions) and one at Chapman University (32 participants, 13 institutions) — specifically to catalogue how research administration software and AI tools are being used inside research offices. The findings map closely onto three workflow areas:

    • Pre-award proposal and budget checking. Institutions are using AI to review draft proposals and budgets for items that will trigger downstream review — facilities requirements, human-subjects protocols, or budget lines inconsistent with a sponsor’s rules.
    • Risk-based compliance screening. AI is used as a first-pass filter that flags transactions, contract clauses, or expenditures for human review rather than replacing that review — described by workshop participants as “an extra layer” that directs attention, not a decision-maker.
    • Contract and reporting automation. Redlining of routine contract language, drafting of progress narratives, and identification of funded projects with commercialisation potential are the most cited post-award use cases.

    Two concrete examples illustrate the pattern at very different institutional scales. Southern Utah University, a smaller teaching-focused institution, built a budget-availability report that automatically flags high-risk expenditures for review — a narrow, operationally specific tool rather than a platform. At the University of California San Diego, a large research-intensive institution, the contracts and grants office is running risk-based proposal review to identify projects needing facilities or IRB attention, and has automated non-disclosure-agreement redlining in a way staff estimate cuts drafting time by roughly 70 percent.

    Workflow stage AI use case Maturity in 2026 Reported example
    Pre-award Proposal/budget risk flagging Early production UCSD risk-based proposal review
    Pre-award Funding-opportunity matching Experimental Faculty-to-grant matching pilots
    Compliance Contract clause / NDA redlining Early production UCSD NDA redlining
    Compliance Expenditure anomaly flagging Pilot Southern Utah University budget-availability tool
    Post-award Progress-report drafting Experimental Institution-reported pilots, Ithaka S+R 2026
    Institution-wide Policy Q&A chatbot for staff Early production UCSD TritonGPT; Emory ORAgpt proof-of-concept

    Evidence From the Field: What Institutions Report

    Two enterprise-level projects sit ahead of the field. TritonGPT, developed at UC San Diego and trained on institutional policy documents, has been available to the campus community since 2023 and is now offered as software-as-a-service to other institutions. At the University of Idaho, the NSF GRANTED-funded AI4RA initiative is building open-source tools for research administrators. At the system level, the California State University system ran a 94,000-response AI sentiment survey — described as the largest of its kind — to set baseline metrics before committing to further rollout.

    These are not isolated enthusiasm projects. The Council on Governmental Relations (COGR) has documented that the U.S. federal government issued more than 200 new or revised policies affecting research administration over the preceding ten years — a compliance burden that is the actual driver behind AI adoption, not novelty. Emory University’s sponsored-programmes office built a proof-of-concept generative AI chatbot, reported by SRA International in May 2025, intended to give research administrators instant, policy-grounded answers rather than requiring them to search static guidance documents.

    Answer-First Q&A

    What is AI actually used for in research administration?

    Institutions report using AI mainly for risk-based screening: flagging proposal budgets, contract clauses, or expenditures that need human review, plus drafting routine reports and answering staff policy questions. It is deployed as a triage layer, not as an autonomous decision-maker in compliance-sensitive workflows.

    Is AI reliable enough for research compliance work?

    Not on its own. Workshop participants in the Ithaka S+R study described current tools as error-prone for high-stakes compliance decisions, so institutions keep a human reviewer in the loop and use AI outputs as a prioritisation signal rather than a final determination.

    What is electronic research administration (eRA) software?

    Electronic research administration (eRA) software centralises pre-award proposal development, post-award financial tracking, IRB/IACUC compliance management, and reporting in one system. Vendors including Cayuse, InfoEd Global and Streamlyne are now embedding AI features into these existing platforms rather than institutions building AI tools separately.

    Will AI replace research administrators?

    Current evidence points the other way. Institutions describe AI as freeing staff time for relationship-building and strategic work, while raising a genuine concern: if entry-level document review and compliance checks are automated away, the profession may lose the training ground that builds administrator judgement over time.

    What Remains Experimental — and Why

    Effort-report anomaly detection — using AI to flag inconsistencies in how research staff certify time charged to federal awards — is frequently proposed as a logical extension of risk-based screening, but publicly documented institutional deployments remain scarce as of mid-2026. This gap matters: effort reporting sits inside some of the most tightly regulated financial-compliance territory in federally sponsored research, and institutions appear to be moving deliberately rather than rushing tools into that specific workflow.

    Three barriers recur across every institution surveyed in the Ithaka S+R workshops:

    • Data governance. Fragmented, inconsistent institutional data undermines AI output quality, and grant proposals routinely contain data covered by HIPAA, export-control rules, or pre-publication intellectual property.
    • Fragmented adoption. Most institutions have not articulated an institution-wide AI strategy for research administration; use is left to individual staff discretion, producing uneven, hard-to-scale experimentation.
    • Trust. Faculty scepticism about whether proposal or compliance data will be used to train external vendor models directly affects whether research administrators can deploy AI tools without damaging working relationships they depend on.

    Implications for Institutions and the Profession

    The practical pattern for institutions considering AI in grants management and compliance workflows is narrower and more disciplined than vendor marketing implies: start with a specific, bounded use case — budget flagging, contract redlining, a policy-guidance chatbot — evaluate it against defined return-on-investment questions, and keep a human reviewer accountable for the final determination. The institutions cited above succeeded by treating AI as an attention-directing layer inside existing research administration workflows, not as a replacement for the judgement that compliance work requires.

    For the broader field of research management and administration, the open question the Ithaka S+R researchers themselves flag is workforce development: if AI absorbs the entry-level document review that has historically trained new research administrators, institutions will need to redesign how professional judgement is built, not just how workloads are reduced. Organisations such as NCURA, SRA International and NORDP are already the venues where this cross-institutional knowledge-sharing is happening, ahead of any formal standard for AI use in the field.

    CASRAI’s own coverage of research administration software categories and standards tracks how eRA platforms are evolving as AI features are absorbed into existing pre-award, post-award and compliance modules — the practical mechanism by which most institutions will encounter AI in this space, rather than through bespoke in-house builds.

  • Electronic Research Administration: What to Evaluate in 2026

    What electronic research administration actually means

    Electronic research administration (commonly abbreviated ERA, and sometimes called eRA) refers to the digital systems and workflows that universities, hospitals, and research institutes use to manage the full lifecycle of sponsored research — from identifying a funding opportunity through proposal submission, award negotiation, compliance monitoring, and financial closeout. The term covers both the specific federal touchpoints, such as the US National Institutes of Health’s eRA Commons and ASSIST systems, and the broader category of institutional research administration software that sits between researchers, sponsors, and finance offices.

    Most research-intensive institutions no longer run these processes on spreadsheets and shared drives. They run them through a dedicated electronic research administration system, or a stack of interoperable modules, because sponsors themselves have moved to electronic submission. Grants.gov, the UK’s UKRI Funding Service, and Horizon Europe’s portal all require electronic workflows on the sponsor side; institutional ERA platforms exist largely to feed proposals into — and pull award data back out of — those sponsor systems without duplicate manual entry.

    Core modules: pre-award, post-award, compliance, effort reporting

    Despite different vendor branding, mature ERA platforms converge on a broadly consistent set of functional modules. The table below summarises what each typically covers and where it interacts with external systems.

    Module What it typically covers External touchpoints
    Pre-award Funding-opportunity discovery, proposal development, budget building, internal sign-off routing Grants.gov, UKRI Funding Service, sponsor portals
    Post-award Award setup, budget tracking, subaward management, financial reporting to sponsors Institutional finance/ERP systems
    Compliance Conflict-of-interest disclosure, IRB and IACUC protocol tracking, export-control screening, foreign-component disclosure Institutional COI registers, ORCID iDs
    Effort reporting Certifying personnel time charged to sponsored awards against actual effort HR/payroll systems, 2 CFR 200.430
    Analytics/reporting Portfolio dashboards, proposal-to-award conversion, audit-readiness reporting Institutional data warehouses

    Few institutions run all five modules from a single vendor. Chief research officers most often report assembling a stack — a proposal-routing tool from one vendor, a dedicated compliance or effort-reporting module from another — connected through system-to-system integrations rather than buying one suite outright. That reality should shape how any evaluation is scoped: interoperability matters as much as feature breadth.

    Why Uniform Guidance and audit scrutiny are reshaping ERA requirements

    US institutions receiving federal research funding operate under the Office of Management and Budget’s Uniform Guidance (2 CFR Part 200). OMB’s 2024 revisions to that guidance — effective for federal awards issued on or after 1 October 2024 — raised the Single Audit expenditure threshold from $750,000 to $1,000,000 and increased the de minimis indirect cost rate available to institutions without a negotiated rate from 10% to 15%. Both changes alter what an ERA system needs to track and report, and by when.

    • A higher Single Audit threshold shifts more institutions toward risk-based, targeted monitoring rather than a full annual audit — which means ERA compliance modules need to surface exception-based flags, not just generate end-of-year reports.
    • The revised de minimis rate changes how budget and indirect-cost calculations should populate proposal templates by default.
    • Effort reporting remains a perennial audit focus area under 2 CFR 200.430, and reviewers increasingly expect systems to certify effort against documented time-and-attendance data rather than after-the-fact estimates.

    Outside the US, UK and EU institutions face parallel pressure: UKRI’s move to its unified Funding Service and Horizon Europe’s stricter foreign-funding disclosure rules both push institutions toward systems that can evidence compliance on demand rather than reconstruct it retrospectively. An ERA platform selected in 2026 needs to be configurable against a moving regulatory baseline, not just the rules in force at implementation.

    A buyer’s framework: what to evaluate before selecting a platform

    Selection committees — typically a chief research officer, sponsored-programs staff, IT, and finance — should evaluate candidate platforms against criteria that go beyond a feature checklist:

    • Configuration versus customisation. Configurable, vendor-supported systems require less internal IT investment but less bespoke fit; heavily customised systems demand ongoing internal development capacity and are harder to keep current when a vendor ships updates.
    • Audit and compliance readiness. Ask vendors to demonstrate exception-based compliance flagging (COI, effort variance, subrecipient risk), not only static reports generated after the fact.
    • Interoperability. Confirm documented integrations with sponsor systems (Grants.gov, eRA Commons, UKRI Funding Service), identity systems (ORCID), and the institution’s own ERP/HR platforms.
    • Total cost of ownership. Homegrown and heavily customised builds frequently carry hidden maintenance costs beyond the initial development estimate; request a multi-year cost breakdown, not just licence price.
    • Vendor stability and support. Research administration software has consolidated significantly through vendor mergers and rebrands over the past decade; ask about implementation timelines, support SLAs, and product roadmap commitments in writing.

    What is electronic research administration?

    Electronic research administration is the use of digital systems to manage the sponsored-research lifecycle — proposal development, award setup, compliance tracking, and financial reporting — in place of paper-based processes. It replaces manual routing and signatures with system-based workflows that connect directly to sponsor submission portals such as Grants.gov.

    What does a research administrator do?

    A research administrator develops and oversees research proposals, awards, and financial transactions on behalf of an institution and its principal investigators. Core duties include budget development, compliance monitoring, and maintaining records that satisfy both institutional policy and sponsor requirements — increasingly through an electronic research administration system rather than paper files.

    What is the difference between eRA and NIH?

    eRA (the NIH’s Electronic Research Administration platform, including eRA Commons and ASSIST) is the online interface through which grant applicants, grantees, and NIH staff exchange administrative information about federal grants. NIH is the funding agency itself; eRA is one agency’s specific electronic system, not a synonym for the broader ERA software category institutions purchase.

    What are ERA systems?

    ERA systems are institutional software platforms — commercial or, less commonly, homegrown — that manage sponsored-research workflows end-to-end. They typically combine pre-award, post-award, compliance, and effort-reporting modules, and connect to external sponsor and identity systems such as Grants.gov and ORCID.

    Implications for institutions, funders, and publishers

    For institutions, the practical implication of tighter Uniform Guidance thresholds and rising audit scrutiny is that ERA selection is no longer purely an IT or finance-office decision — it is a compliance-risk decision that belongs on the chief research officer’s desk. Systems chosen primarily on price or user-interface polish, without a documented compliance-flagging capability, risk becoming an audit liability rather than an efficiency gain.

    For funders and publishers, the growth of ERA adoption strengthens the case for standardised metadata at the point of proposal and award creation — identifiers such as ORCID iDs and the Research Organization Registry (ROR) reduce downstream reconciliation work when award data eventually needs to map to publications, contributor roles, and institutional affiliations. Professional bodies including NCURA, ARMA, EARMA, and INORMS have each published guidance and community benchmarking on ERA adoption, reflecting how central this tooling decision has become to the research-administration profession globally.

    Outlook: ERA selection as a 2026 strategic priority

    The direction of travel is clear: sponsors are tightening disclosure and audit expectations at the same time as institutions face budget pressure to do more with fewer administrative staff. An ERA platform that cannot demonstrate compliance readiness against a moving regulatory baseline — and that cannot interoperate cleanly with sponsor and identity systems — will struggle to justify its cost within two to three budget cycles. Institutions evaluating platforms in 2026 should treat the selection process as an ongoing compliance investment rather than a one-off procurement exercise, revisiting vendor roadmaps annually against the next round of Uniform Guidance and sponsor-portal changes.

    Institutions building out their research administration function more broadly can also consult CASRAI’s research administration resources and the CASRAI Dictionary for grounded definitions of the compliance and reporting terms that ERA systems are built to track.