Skip to Content
Free evaluation matrix · 25 criteria · vendor-neutral scoring

Score every AI-in-ERP vendor on the same 25 lines.

Comparison calls blur together because every vendor demos their best five minutes. This matrix makes them all answer the same 25 questions across four categories (capability, governance, data handling, cost), each scored 0–3 with a weight you set. The full matrix is on this page; the editable, auto-totaling sheet is yours for an email.

Capability: can it do real work? (8 criteria)

Score against your workflows on your data, never against the demo.

#CriterionWhat a 3 looks likeWeight
C1Answers cite source recordsEvery number links to the record it came from, one clickhigh
C2Reads live dataLive database, not a synced copy with stalenesshigh
C3Cross-module reachOne question spans sales, stock, accounting in one answerhigh
C4Prepares real work productDraft quote/reminder/transfer as actual records, not chat texthigh
C5Context across a conversationThird related question still knows what the first meantmed
C6Works where your team worksIn your chat, on your screens, in your AI toolsmed
C7Version & edition coverageYour Odoo version and edition, today, no upgrade requiredhigh
C8Handles your messy dataGraceful on ambiguity; says so when unsuremed

Governance: who is in control? (7 criteria)

The category where demos are quietest and incidents are loudest.

#CriterionWhat a 3 looks likeWeight
G1Acts under asker's permissionsNo service-account superuser; your access rights decidehigh
G2Write policy as settingsAllow/hold/forbid configurable per model and actionhigh
G3Approval before commitDrafts held; a person triggers every send/posthigh
G4Value caps & rate limitsAmount thresholds and write ceilings you setmed
G5Refusals loggedOut-of-scope requests produce visible trail rowshigh
G6Full decision trailOne row per decision, named person, exportable alonehigh
G7Scoped multi-assistantPublic-facing bot provably can't reach the ledgermed

Data handling: where does it go? (5 criteria)

Get every answer in writing; a vendor without one has answered.

#CriterionWhat a 3 looks likeWeight
D1No training on your dataContractual, not FAQ-levelhigh
D2Egress transparencyField-level list of what leaves per requesthigh
D3Redaction before modelsSensitive fields stripped at egressmed
D4Retention & deletionStated windows and a deletion path, in writingmed
D5Residency / on-prem pathClear answer on where processing happens, options if it mattersmed

Cost & commitment: what does it really cost? (5 criteria)

Price the three-year picture, never the first invoice.

#CriterionWhat a 3 looks likeWeight
$1Transparent pricingPublic model; no quote-only mysterymed
$2Pilot on your dataProve-it phase on your instance before commitmenthigh
$3No per-module surprisesScope of tools included is explicitmed
$4Exit costYour data and configs leave with youmed
$5Renewal story in writingWhat changes at renewal, stated nowmed
Scoring: 0 = absent · 1 = claimed, not shown · 2 = shown on demo data · 3 = shown on your data. Weight high = ×2. Disqualifiers regardless of total: a zero on G1, G3, G6 or D1. The spread between vendors' totals matters less than where the zeros land. Zeros cluster where incidents will.
the editable sheet

Take the matrix into your evaluation.

The sheet has all 25 criteria with auto-totaling weighted scores, a column per vendor, and a notes field per line, ready for your selection file. We send it once. No marketing emails unless you opt in later.

Prototype form; when enabled, we store your details in our Odoo CRM to send the sheet and respond if you reply. No marketing emails unless you opt in; see our privacy policy.