Dynamics SL + on-prem AI
AI for Dynamics SL, grounded on your project and contract data
Short answer
Dynamics SL still runs project accounting and DCAA-compliant timesheets for a lot of government contractors who have no plan to move off it soon. A private LLM grounded read-only on SL's SQL Server database can answer project, contract, and indirect rate questions in plain language, and can run entirely on infrastructure a CIO controls, without sending CUI or contract cost data to a public AI API.
- ERP
- Microsoft Dynamics SL, Dynamics SL 2018, Solomon IV (legacy)
- Industries
- Government Contracting, Aerospace, Defense, Professional Services
- Written for
- CIO
Dynamics SL is a project and contract accounting system built for exactly the kind of organization Microsoft has mostly stopped talking about: government contractors, professional services firms, and aerospace and defense subcontractors who need DCAA-compliant timesheets, indirect rate pools, and funded-value tracking down to the task order line. Microsoft's newer investment goes to Business Central and Dynamics 365 Project Operations, but a large population of SL customers have no near-term reason to move, because the system does exactly the specific job it was built for.
The problem is not that SL fails at its job. It is that nobody outside the controller's office and a couple of finance staff can get a straight answer out of it without a report request or a spreadsheet pull. A program manager asking what is left on a task order's funding ceiling, or an indirect rate group asking why the provisional overhead rate moved this period, waits on the same two or three people every time, and those people are also the ones preparing the annual incurred cost submission.
That is a grounding problem, not a system-replacement problem. SL's project and contract data lives in an ordinary SQL Server database, reachable read-only, and an AI layer built on a proper semantic mapping of its Project Controller, Contract, and Timesheet tables can answer the recurring questions directly, cite the source records, and leave the actual DCAA labor-charging controls and approval workflows exactly as they are.
For a CIO evaluating this, the constraint that matters most is data handling. Dynamics SL customers who are DCAA-audited or CMMC-scoped cannot casually route contract cost and labor data through a shared public AI service. The architecture that fits is the same one that fits any regulated ERP: read-only connection to SL's database, a model that runs on hardware the company controls, and a design that never writes back to a labor or billing record without a person approving it first.
What usually gets in the way
The problems we hear most from cio teams running Microsoft Dynamics SL.
DCAA audit prep is a manual scramble every year
Building the incurred cost submission means pulling project ledger detail and mapping it to ICE schedules by hand, usually under deadline pressure and usually by the same one or two people.
Indirect rate variance takes a spreadsheet rebuild every close
Explaining why the provisional overhead or G&A rate moved this period means recreating the rate pool calculation manually, since SL does not narrate the variance for you.
Funded value and burn-rate risk surfaces late
A task order approaching its funding ceiling is often discovered when billing is about to exceed it, not weeks earlier when a program manager could still act.
Legacy VB.NET customizations are undocumented
Custom screens built with the Dynamics SL SDK years ago run the business today, but the people who wrote them have often moved on, leaving no real documentation.
Microsoft's own AI roadmap has bypassed SL
Copilot and generative AI investment goes to Business Central and Project Operations; SL customers get no native path to the same capability inside the product.
Where AI earns its place in Microsoft Dynamics SL
Each use case names the ERP objects it reads or writes, so your ERP team can judge the integration effort before anyone commits budget.
Plain-English project and contract status
A program manager asks what has been billed against a contract, what remains on a funded task order, or where a project stands against budget, answered from live Project Controller and Contract module data.
Touches: Project Controller (project ledger, budget), Contract module (funded value, billed value, ceiling)
Outcome: Cuts routine status questions from a request to the PM office down to a self-service question with the source records cited.
DCAA timesheet exception review
An agent reviews recent Timesheet Entry and labor distribution activity for missing entries, late submissions, and unusual charging patterns ahead of a compliance review, rather than waiting for an audit sample to find them.
Touches: Timesheet Entry, Labor Distribution, employee/project charge codes
Outcome: Surfaces labor-charging exceptions before an auditor does, without changing SL's own timesheet approval workflow.
Indirect rate variance explanation
A controller asks why the provisional overhead or G&A rate changed this period and gets a plain-language walk-through of the pool and base changes driving it, grounded in the actual rate pool tables.
Touches: Overhead/G&A cost pools, provisional billing rate tables, GL account groupings
Outcome: Turns a spreadsheet rebuild into a grounded explanation delivered in minutes, with every figure traceable to a GL account.
Incurred cost submission prep assistant
Pulls project ledger detail and maps it to the relevant ICE schedule categories, giving the accounting team a structured starting draft instead of a blank spreadsheet each year.
Touches: Project ledger detail, GL account mapping, indirect cost pool structure
Outcome: Shortens the annual ICE/incurred cost submission prep cycle, with the underlying transactions cited for the audit trail.
Funding ceiling and burn-rate early warning
A scheduled agent compares billed and committed cost against each task order's funded value and flags orders trending toward their ceiling before it becomes a stop-work risk.
Touches: Contract module funded/billed/ceiling values, Project budget
Outcome: Gives program managers weeks of lead time instead of discovering a ceiling breach at invoice time.
AP document intake and PO matching
OCR and classify vendor invoices, then propose a match against open purchase order lines in SL for a clerk to confirm before it posts.
Touches: Accounts Payable module, purchase order detail
Outcome: Cuts manual keying and matching time on routine vendor invoices while leaving posting to a person.
Legacy customization documentation assistant
Point a code-aware model at exported Dynamics SL SDK screens and VB.NET customizations to generate plain-English descriptions of what each one does, ahead of any future migration decision.
Touches: SL SDK custom screens, VB.NET/COM customization source
Outcome: Builds a documentation baseline in weeks instead of depending on the one person who remembers why a screen was built that way.
Reference architecture
A read-only layer on Dynamics SL's SQL Server database, grounded on Project Controller, Contract, and Timesheet data through a purpose-built semantic mapping, with model serving kept off any public AI API for CUI and contract-cost sensitivity.
- 1
SL connectivity
Direct SQL Server read replica or a scoped, read-only database login against the SL database; the Dynamics SL SDK is used only where a specific, approved write-back is later required.
- 2
Semantic layer
A mapping from SL's project, contract, and timesheet table and field names to the vocabulary a program manager or controller actually uses, built with input from whoever runs project accounting today.
- 3
Model serving
Open-weight models served on infrastructure the company controls, sized to the organization's query volume, with no contract cost or labor data sent to a shared public AI service.
- 4
Retrieval and agents
Text-to-SQL for structured project, contract, and rate questions; scheduled exception agents for timesheet review and funding ceiling monitoring; document RAG for AP and contract file intake.
- 5
Governance and audit
Every answer logged with the specific project, contract, or GL records it drew from, so a controller or DCAA auditor can trace an AI answer back to source transactions.
Integration notes for your ERP team
- Use a scoped, read-only SQL Server login against the SL database, or a nightly/near-real-time replica, rather than granting broad access through the SL SDK.
- Build the semantic layer with the controller or whoever currently owns indirect rate calculations; SL's project and contract table structure is coherent but not self-explanatory to an LLM without that mapping.
- Keep write-back out of scope for the initial rollout; if it is added later (for example, drafting a timesheet correction for review), route it through the SL SDK's standard APIs with explicit human approval.
- Confirm how provisional billing rates and rate pool structures are configured for the specific contracts in scope; rate methodology varies enough between GovCon shops that this cannot be assumed generic.
- If SL customizations built with the SDK are heavily used, export their source for the documentation use case separately from the live database connection, so the model never needs write or execute access to custom code.
- Size funding ceiling monitoring to check against committed cost, not just billed cost, since committed-but-unbilled cost is usually the earlier warning signal a program manager needs.
- Log every generated query and every answer with the source table and record identifiers, since this is exactly the kind of evidence a DCAA or CMMC assessor will want to see if the AI layer is ever in scope for review.
Deployment options
Air-gapped on-prem
Defense subcontractors and CMMC-scoped GovCon firms who cannot let CUI or contract cost data touch a network path outside their control.
Model, semantic layer, and SL read replica all run on company-owned hardware with no outbound path for inference traffic.
Private or sovereign cloud
Contractors comfortable with cloud economics but not with a shared public AI API touching labor and contract cost data.
Dedicated-tenant deployment connecting to SL over a controlled, logged connection, with data residency matched to existing compliance obligations.
Hybrid
Organizations wanting to prove one use case, such as funding ceiling monitoring, before committing budget to a wider rollout.
A scoped pilot on modest hardware, same read-only architecture, extended once the use case has proven its value.
Compliance and data control
How the architecture supports your obligations. Certification and accountability stay with your organisation; the design keeps the evidence straightforward.
DCAA compliance
The AI layer reads timesheet and labor distribution data to surface exceptions; it does not alter labor charging, approval routing, or any control DCAA already audits.
CMMC / DFARS 252.204-7012 / NIST SP 800-171
Where SL data touches CUI, model serving and the semantic layer stay on infrastructure inside the same enclave, with no data sent to a public AI API.
ITAR / export control
For defense-adjacent contractors, technical and contract data stays on company-controlled hardware, with access scoped to authorized US persons where required.
GAAP and internal financial controls
Rate variance and incurred cost explanations are read-only aids to the controller; the actual journal entries and rate calculations remain SL's system of record, unaltered by the AI layer.
Where Netray fits
Custom build
DCAA timesheet exception logic, incurred cost submission mapping, and funding ceiling monitoring are specific enough to GovCon project accounting that they are typically built as a custom layer on top of SL's data.
ERPray
Once the SL semantic layer exists, ERPray's connector pattern can provide general-purpose, cited question-answering over project and contract data for program managers and finance staff.
How an engagement runs
Phase 1 . 2-3 weeks
Discovery
- -Project, Contract, and Timesheet table mapping to business vocabulary
- -Confirmed read-only connectivity approach
- -Security review with IT and the controller's office
- -Prioritized use case and success criteria
Phase 2 . 6-8 weeks
Pilot
- -Working semantic layer for the pilot data domain
- -Model serving stood up on company-controlled infrastructure
- -5-10 real project or contract questions answered end to end with cited sources
- -Pilot review with CIO and controller
Phase 3 . 8-10 weeks
Production
- -Hardened replica or connection with defined refresh cadence
- -Role-based access aligned to existing SL security
- -Audit logging suitable for DCAA or CMMC review
- -Runbook for IT to operate and maintain the connector
Phase 4 . ongoing
Scale
- -Additional domains (AP intake, rate variance) added to the semantic layer
- -Funding ceiling monitoring extended across all active contracts
- -Quarterly review of query patterns and audit-readiness
Questions to ask any vendor, including us
A short list that separates real Microsoft Dynamics SL AI work from a chatbot demo.
- Does your approach require any change to SL's timesheet or labor-charging controls, or is it strictly read-only?
- Where does the model run relative to our network, and can you guarantee no contract cost or CUI data reaches a public AI API?
- How do you build the mapping between SL's project and contract tables and our actual DCAA rate methodology?
- Who builds and maintains the semantic layer, and how much of our controller's time does that require?
- Can you show a real example of the SQL your system generated against Project Controller or Contract module data?
- How does this hold up under a CMMC or DCAA audit; what evidence trail does the AI layer produce?
- What happens to this project if we later migrate off Dynamics SL?
- What is your rollback plan if the read replica falls behind during a close or an audit period?
Frequently asked questions
Can AI answer project and contract questions directly from Dynamics SL?
Yes. SL's project, contract, and timesheet data lives in an ordinary SQL Server database, reachable read-only. Once a semantic layer maps that data to plain business terms, an AI layer can answer status, funding, and rate questions with the source records cited, without changing anything in SL itself.
Does this change or automate DCAA labor charging?
No. The standard design is read-only: it reviews timesheet and labor distribution data to flag exceptions for a person to review, and it does not alter labor charging, approval routing, or any control DCAA already audits.
Is our contract cost and CUI data safe from a public AI API?
It can be kept entirely off one. Model serving, the semantic layer, and the SL read connection can all run on infrastructure your organization controls, which is the standard architecture for CMMC-scoped or ITAR-relevant contractors.
Can this help with the annual incurred cost submission?
Yes. An agent can pull project ledger detail and map it to the relevant ICE schedule categories, giving accounting a structured draft with every figure traceable to source transactions, rather than starting from a blank spreadsheet.
Will this catch a task order approaching its funding ceiling?
A scheduled monitoring agent can compare committed and billed cost against each task order's funded value and flag orders trending toward their ceiling well before billing would exceed it, giving program managers time to act.
What about our old SL SDK customizations nobody documented?
A code-aware model can read exported VB.NET and SDK customization source and generate plain-English descriptions of what each screen does, useful as a documentation baseline for onboarding staff or scoping a future migration, without executing or modifying the code.
Does this require us to plan a migration off Dynamics SL?
No. The architecture is designed to add value to the system you are running today. It also happens to produce useful documentation and usage data if a migration decision comes up later, but that is not a prerequisite.
Related guides
ITAR-Compliant AI for ERP Technical Data
How to add generative AI to your ERP without creating a deemed export under ITAR. On-prem architecture patterns an Empowered Official can sign off on.
DFARS 7012 + NIST 800-171Mapping DFARS 7012 and NIST 800-171 Controls to AI on Your ERP
Map DFARS 252.204-7012 and NIST SP 800-171 control families to an AI system layered on your ERP, with evidence a Compliance Manager can defend.
GovCon accounting + DCAA + on-prem AIAI for Government Contractor ERP Under DCAA Scrutiny
AI for government contractor ERP under DCAA scrutiny: draft indirect rate variance narratives, flag unallowable costs, and build audit trails, run on infrastructure you control.
ERP AI Buyer GuideHow to Choose an ERP AI Implementation Partner
A CIO checklist for picking an ERP AI implementation partner: the architecture questions to ask, red flags, pricing models, and what to demand in the SOW.
RAG + SQL + permissionsA Private LLM Grounded on Your ERP Data
How a private LLM answers questions on your ERP data: RAG plus text-to-SQL, role-based permissions inherited from the ERP, and where each fits.
D365 F&SCM + private AIAI for Dynamics 365 Finance and Supply Chain beyond Copilot
AI for D365 Finance and Supply Chain beyond Microsoft Copilot: private LLM grounded on data entities and OData, on-prem-capable via Local Business Data.
Plan it with numbers
DCAA Accounting Compliance Readiness Assessment
Score your accounting system against the SF 1408 preaward survey criteria and DCAA audit expectations across timekeeping, job costing, and indirect rates.
Free ToolGovernment Contract Compliance Cost Calculator
Estimate the implementation, assessment, and recurring annual cost of DFARS/CMMC compliance for your DoD business, and see it as a percentage of contract revenue.
Free ToolERP AI Maturity Assessment
Benchmark how deeply AI and automation are embedded in your ERP operations, from data foundations to autonomous agents, across four maturity levels.
GuideOn-Prem AI for Defense Contractors: The Complete Guide
On-prem AI for defense contractors: deploy LLMs and AI agents inside your CMMC and ITAR boundary. Architecture, hardware costs, timelines, and vendor options.
GuideThe CFO Guide to AI ROI in Manufacturing
A practical CFO guide to AI ROI in manufacturing: payback benchmarks, cost models, and how to separate real returns from vendor hype before you sign.
GuideERP Vendor Selection & Evaluation Scoring Matrix
Use a weighted scoring matrix to evaluate ERP vendors objectively. Compare functionality, TCO, implementation risk, and vendor viability across platforms.
Talk it through with an engineer who knows Microsoft Dynamics SL
Bring one real question your team cannot answer from the ERP today. We will map the data path, the model, and where it runs, and tell you honestly if AI is the wrong tool for it.