# Accounts payable automation: what twelve AI models recommend, and why, September 2026

Finance AI Recommendation Index, September 2026 Edition. Page: https://finance-ai-index.com/spend/accounts-payable-automation/guide/

Twelve AI models were asked for accounts payable automation software six ways each, on behalf of a small, a mid-market and an enterprise B2B company: 216 answers, in which a judge labeled 60 products. Seven of them carry at least 10 labels and a first choice, and are ranked. This guide walks the top seven in the order the mid-market standing puts them, with what the models said for and against each, in their words. Stampli leads at 19% of first choices, and the standing is contested.

## At a glance

| # | Product | Share | Negative rate | Labels |
|---|---|---|---|---|
| 1 | Stampli | 19% | 0% | 37 |
| 2 | BILL | 19% | 23% | 52 |
| 3 | Melio | 19% | 6% | 17 |
| 4 | Ramp | 16% | 10% | 42 |
| 5 | AvidXchange | 12% | 16% | 38 |
| 6 | Tipalti | 7% | 11% | 44 |
| 7 | Medius | 5% | 0% | 16 |

## 1. Stampli

Stampli is #1 of 7 at 19% of first choices, 37 labels, negative rate 0%. By buyer: small business unranked; mid-market #1 of 7 at 19%; enterprise #9 of 9 at 2%.

- For: "I'd start with Stampli. It's the strongest "classic AP automation" choice for mid-market companies" (GPT-5.4 mini, direct prompt)
- For: "Stampli (best for ERP integration and conversational AI that actually works on messy invoices)" (DeepSeek V4 Flash, negative prompt)

## 2. BILL

BILL is #2 of 7 at 19% of first choices, 52 labels, negative rate 23%. By buyer: small business #1 of 4 at 41%; mid-market #2 of 7 at 19%; enterprise unranked.

- For: "BILL AP/AR and AvidXchange are the top recommendations due to their strong user ratings, scalability, and comprehensive feature sets." (Mistral Small, paraphrase prompt)
- Against: "Verdict: Avoid if you have complex approval workflows, rely on tight cash flow, or have a low tolerance for support friction." (DeepSeek V4 Flash, negative prompt)

## 3. Melio

Melio is #3 of 7 at 19% of first choices, 17 labels, negative rate 6%. By buyer: small business #3 of 4 at 20%; mid-market #3 of 7 at 19%; enterprise unranked.

- For: "Recommendation: Start with Melio if payments are simple/low-volume and you want zero commitment." (Grok 4.1 Fast, budget prompt)
- For: "For the tightest budgets, start with Melio's free plan or Ramp's free tier to test before committing." (Claude Haiku 4.5, budget prompt)

## 4. Ramp

Ramp is #4 of 7 at 16% of first choices, 42 labels, negative rate 10%. By buyer: small business #2 of 4 at 27%; mid-market #4 of 7 at 16%; enterprise #7 of 9 at 5%.

- Against: "Verdict: Avoid if AP automation is your primary need." (DeepSeek V4 Flash, negative prompt)

## 5. AvidXchange

AvidXchange is #5 of 7 at 12% of first choices, 38 labels, negative rate 16%. By buyer: small business unranked; mid-market #5 of 7 at 12%; enterprise unranked.

- For: "AvidXchange | Mid‑sized businesses wanting a complete, cloud‑based AP solution ... General mid‑market AP automation (no global payments) | AvidXchange, BILL, Stampli" (GLM 4.7 FlashX, paraphrase prompt)
- For: "AvidXchange | Mid‑market B2B (100–1,000 employees) | Deep focus on mid‑market AP; strong ERP integrations; B2B payment network" (GLM 4.7 FlashX, direct prompt)
- Against: "AP) automation tools you should approach with caution, along with the primary reasons why users often recommend avoiding them.

### 1. AvidXchange" (Qwen 3.7 Flash, negative prompt)
- Against: "Significant vendor-side and operational concerns ... BBB rating 1.14/5 stars (37 reviews) — extremely poor" (Kimi K2, negative prompt)

## 6. Tipalti

Tipalti is #6 of 7 at 7% of first choices, 44 labels, negative rate 11%. By buyer: small business unranked; mid-market #6 of 7 at 7%; enterprise #1 of 9 at 18%.

- For: "If you want a single default recommendation, I'd start with Tipalti for a typical mid-market B2B company" (Perplexity Sonar, direct prompt)
- Against: "This is a cash-flow killer and was a major downfall for tools like Tipalti and Airbase." (Qwen 3.7 Flash, negative prompt)
- Against: "Reddit users strongly advise avoiding it." (Grok 4.1 Fast, negative prompt)

## 7. Medius

Medius is #7 of 7 at 5% of first choices, 16 labels, negative rate 0%. By buyer: small business unranked; mid-market #7 of 7 at 5%; enterprise #5 of 9 at 10%.

- For: "Medius - provides AP automation for midsize businesses, removing manual AP tasks, strengthening controls, and gaining real-time visibility" (Llama 4 Maverick, paraphrase prompt)
- For: "The top AP automation software solutions in 2026 are Medius, Tipalti, Stampli, and AvidXchange." (Llama 4 Maverick, direct prompt)

## Questions the record answers

**Which accounts payable automation software do AI models name first most often?** Stampli, in 19% of first choices for a mid-market B2B company in the September 2026 Edition, from 37 labels. The standing is contested: the models did not settle on one product.

**Does the answer change with the size of the company?** Small business: BILL at 41%. Mid-market: Stampli at 19%. Enterprise: Tipalti at 18%. Each standing is computed within its segment and never pooled.

**Which framing changes the answer?** The direct, paraphrase, budget and scale framings count toward share. The product named first differs by framing: direct AvidXchange; paraphrase BILL; comparative Ramp; budget-constrained Melio; negative Stampli. The table above has the counts.

**How stable is one answer?** Asked again with nothing changed, the models moved their own first choice 68% of the time across the repeat sample; This category was in the repeat sample: asked again with nothing changed, the leader's share moved 0 points and the leader changed to Melio.

**Where do the answers come from?** Twelve of the twelve models return sources. Their 66 answers here cite 782 pages; the sites cited most are ramp.com, learn.g2.com, medius.com. The category page lists them all.

**Can a vendor pay to be on this list?** No. A product is on the record because a model named it. A vendor can claim its page, propose corrections to the vendor table and be told when its standing moves; it cannot change a label, a share or a rank, and the publisher's conflicts are disclosed on the method page.

Published under CC BY 4.0. The output is the models' output; nothing here is a recommendation by the index.