From Ritu Raj | Product & Market Analysis
Enterprise Sales Cycles Stretch as AI Buying Committees Double to 14 People
On this page
Forrester's 2026 buyer research found that when a purchase includes generative AI features, the buying group doubles, from 7 members to 14. That is the cleanest number available on why the enterprise sales cycle for AI products feels longer. The new seats belong to security, legal, privacy, AI governance and finance, and each one arrives with its own review. Nobody publishes an AI-only cycle length in days, so this post works from the reviewers instead.
Key takeaways
- GenAI features double the buying group. Forrester's survey of nearly 18,000 business buyers found 14 members on purchases with genAI features against 7 on purchases without them.
- The delay now arrives after the buyer has already picked you. G2's 2026 survey of more than 1,000 software buyers found IT security review is the single biggest source of delay after vendor selection, cited by 39% of buyers and 50% of enterprise buyers.
- Finance is the reviewer most likely to reverse a yes. Finance involvement in software decisions rose from 31% to 46% in one year, and 49% of buyers said their CFO reversed a deal the buying team had already approved.
- No public dataset times AI deal cycles in days. The evidence is structural: more reviewers, stricter AI evaluation rules, and Ebsta data showing win rates fall 67% when deals slip. Treat any single AI cycle-length figure with suspicion.
What changed in the enterprise sales cycle for AI products
An enterprise sales cycle is the elapsed time from a qualified opportunity to a signed contract. For most of the last decade its shape was stable. A business owner found a problem, a champion ran an evaluation, procurement negotiated, and IT checked the integration.
AI products broke that shape in a specific way. They process data the buyer considers sensitive, they produce output nobody can fully predict, and they are often priced on usage that finance cannot forecast. Each of those properties belongs to a different department, and each department now wants its own review.
Short answer: AI deals take longer because a purchase with genAI features brings roughly twice as many people into the buying group, and the added people are reviewers rather than users. Security, legal, privacy, AI governance and finance each run a separate check, usually after the business owner has chosen a vendor, and usually one after another.
The result is not one longer review. It is several new ones, owned by people the seller never met during discovery. G2's 2025 survey of 1,169 B2B decision-makers found that about 8 in 10 buyers now face stricter requirements for AI software evaluations from IT security, legal and compliance teams.
Getting onto the vendor list in the first place is a separate problem, covered in how buyers now build the AI shortlist before contacting sales. This post starts where that one ends: the buyer has picked you, and the reviews begin.
The B2B buying committee doubled for genAI purchases
A buying committee, which Forrester calls a buying group, is everyone who influences or signs off on a purchase. Forrester's State of Business Buying 2026, summarised by VP and principal analyst Barbara Winters in January 2026, puts the average at 13 internal stakeholders and nine external participants.
The genAI split is the more useful number. For purchases that include genAI features, the buying group doubles to 14 members, against 7 without those features. The report rests on Forrester's Buyers' Journey Survey of nearly 18,000 global business buyers, fielded in 2025.
Forrester's public summary does not reconcile the 14 with the 13 plus 9, and the two figures appear to count different things. Do not add them together. Read the 14 against the 7, which is a like-for-like comparison inside one survey.
Procurement is the other structural change. Forrester found procurement professionals act as decision-makers in 53% of buying cycles, and the C-suite was cited by 68% of respondents, according to Digital Commerce 360's coverage of the same report. Procurement as a decider rather than a processor changes what the final month of a deal looks like.
Buyers do not see the bigger group as a problem, which is worth knowing before you complain about it. Among respondents with six or more internal participants, 94% said larger groups delivered clear benefits. Fewer than half cited slower decision-making as a significant drawback.
My read is blunt. You are not going to talk a buyer out of the committee, and sellers who try sound like they are hiding something. The committee is the customer now, and the job is to sell to every seat in it.
Enterprise sales cycle length: what the data actually shows
Here is the uncomfortable part. No vendor-neutral source found for this post publishes an AI-specific enterprise sales cycle length in days with a stated method. Plenty of pages quote one. The ones checked here traced back to unnamed data or to tools that do not disclose a sample.
What does exist is a set of general benchmarks with methods attached. They agree on direction far more than they agree on numbers, and the direction is the useful part.
Vendor selection is no longer the finish line
G2's 2026 Buyer Behavior Report surveyed more than 1,000 B2B software buyers and interviewed more than 50 sales and marketing leaders. It found evaluation is now the longest stage of the buying journey. It overtook research for the first time, and 40% of buyers named it the longest stage.
The delay sits after the choice is made. Once a vendor is selected, IT security review is the single biggest source of delay, cited by 39% of buyers overall and 50% of enterprise buyers. Budget approval at 32% and implementation planning at 25% follow close behind.
That ordering matters for how you forecast. A deal at "vendor selected" used to be a deal in paperwork. On an AI deal it is a deal entering its riskiest phase, and most CRM stage definitions still treat it as nearly closed.
What a slipped deal costs
Ebsta and Pavilion's 2024 benchmarks analysed 4.2 million opportunities from 530 companies, representing over $54 billion in revenue. Sales cycles grew 16% in the first half and ran 38% longer than in 2021, before shortening by 23% as the year closed. Treat that as a general market reading, not an AI one.
The more durable finding is about slippage. When deals slipped, win rates fell by 67%, particularly for those delayed over 8 weeks, and 44% of deals were pushed back. More than 7 days of inactivity with no future activity booked cut win rates by 65%.
For a baseline in days, Gong's State of Revenue 2025, as summarised by SaaStr's Jason Lemkin, puts $100,000 deals at about 70 days to close. The summary does not state the number of deals or the period covered, so treat it as directional. It describes SaaS deals in general, not AI deals.
Put those together and the mechanism is visible even without an AI cycle figure. Each added reviewer is another chance for a deal to go quiet for a week. A quiet week is exactly what the Ebsta data punishes.
Four reviewers an AI deal must clear, and what each needs
The playbook below is organised by reviewer, because that is how the delay is organised. Each reviewer asks a different question, accepts a different document as an answer, and should receive it at a different point in the cycle.
| Reviewer | The question they are paid to ask | What usually stalls it | What to send, and when |
|---|---|---|---|
| Security | Where does our data go, and can your system be attacked through it? | A questionnaire answered from scratch, with the AI sections left vague | An AI security addendum, before the first technical call |
| Legal and privacy | Does our data train your model, and who is liable when output is wrong? | Redlines on training rights and liability caps, opened at contract stage | A training-use clause in the order form, in week one |
| AI governance | Can we explain this system, monitor it and switch it off? | No single owner, so the review waits for a committee meeting | A one-page system description the committee can file |
| Finance | What will this cost at real usage, and when does it pay back? | Usage pricing with no ceiling, then a CFO reversal after approval | A cost model at three usage levels with a payback date |
Security: answer the AI questions before they are asked
Security review is the delay with the strongest evidence behind it, from G2's 39% overall and 50% enterprise figures. The fix is not answering faster. It is answering earlier, before the reviewer has formed a view of you from a half-completed form.
Build an AI addendum to your standard security pack. Cover model providers, where prompts and outputs are stored, retention periods, tenant isolation, and how you test for prompt injection. The audit evidence that holds up under scrutiny is covered in the guide to SOC 2 evidence for AI systems.
One finding cuts against intuition. G2 reports that information security involvement in decisions fell from 32% to 25% over the year, even while security review stayed the biggest delay. Fewer security people are in the room, and the ones who are hold the gate. Find out their names in the first week.
Legal and privacy: settle training rights before the redlines
The IAPP's AI Governance Profession Report 2025 found privacy teams and legal and compliance teams each hold primary ownership of AI governance at 22% of organisations. Your buyer's AI review is as likely to sit with a privacy lawyer as with an engineer.
The privacy lawyer's first question is usually whether customer data trains your model. Put the answer in the order form, not on a policy page the buyer has to go and find. The clauses that come up most are set out in the breakdown of data processing agreement clauses for AI vendors.
I would rather concede a strict no-training clause on day one than discover in week ten that it is the clause holding up signature. A clause you were always going to accept is worth nothing as a late concession and costs weeks as a late negotiation.
AI governance: give the committee something it can file
AI governance is the newest reviewer and the least predictable. The IAPP found 77% of surveyed organisations are working on AI governance, rising to near 90% of those already using AI. The survey covered more than 670 respondents in 45 countries and was fielded in spring 2024, which makes it the oldest figure in this post.
Ownership is scattered. Privacy leads at 22%, legal and compliance at 22%, IT at 17%, data governance at 10%, ethics and compliance at 6% and security at 5%. A scattered owner usually means a committee that meets on a fixed calendar, and a monthly committee is a 30-day delay with an agenda.
Send a short system description the committee can approve without convening. List the models, the data each one sees, where a human reviews output, what is logged, and how the buyer switches a feature off. A working format is in the AI bill of materials disclosure template, and the buyer-side view of the same review is in the AI agent procurement checklist.
Finance: arrive with a payback date
Finance is the reviewer most likely to reverse a decision that already looked won. G2 found finance involvement in software decisions rose from 31% to 46% in one year. In the previous 12 months, 49% of buyers said their CFO reversed a deal the buying team had already approved.
AI budgets sharpen that. In companies with dedicated token or LLM budgets, 54% of buyers reported a CFO veto of an approved purchase, against 29% in organisations without one. Three in four buyers who experienced a late-stage veto now expect positive return within six months of signing.
Give finance a cost model at low, expected and high usage, plus a date by which the deployment pays back. Expect requests for shorter terms as well: vetoed buyers push for contracts under 12 months at 40%, against 18% of other buyers. The terms CFOs now ask for are mapped in the guide to CFO contract clauses for AI purchases.
Run the reviews in parallel, not in series
Most AI deals run their reviews one after another. The champion finishes the evaluation, then hands to security, then legal sees the paper, then finance sees the number. Each handoff is a week where nothing is booked, and the Ebsta data says that week costs win rate.
The alternative is to open every review in the first two weeks. Ask the champion during discovery for the name of the security reviewer, the privacy or legal owner, whoever runs AI governance, and the finance approver. Then send each one their document in the same week.
The chart is a model of the logic, not a measurement. Your reviews will take different lengths of time, and some genuinely depend on each other, because legal may wait for the security findings. Even partial overlap removes the idle weeks between handoffs.
Parallel reviews need someone on your side who can answer technical, legal and commercial questions without a fortnight's turnaround. That is one reason sales engineering keeps absorbing work that used to sit elsewhere, the shift covered in the piece on sales engineering as the new marketing.
My view: the cheapest acceleration available to an AI seller is not a discount. It is a named contact for each reviewer and a dated document in that reviewer's inbox by the end of week two. Discounts shorten the finance review and do nothing for the other three.
Where this argument is weakest
The case above rests on a chain of reasoning, and two links in it are weaker than the rest. Both are stated here so you can weigh them yourself.
The committee may be shrinking, not growing
G2's 2025 survey reported the opposite of Forrester on size. It found once-dominant buying committees of 5 to 8 members shrinking, with smaller groups of 3 to 4 rising to prominence. The two surveys ask different questions of different samples, so both can be partly true, but the direction is not settled.
One reconciliation is a smaller core of deciders plus a larger ring of reviewers who influence without voting. That fits both datasets and fits the security and finance findings above. It is my reading, though, and neither firm has published it.
The AI cycle penalty has not been timed
No source in this post measures how many days AI features add to an enterprise sales cycle. Ebsta and Gong measure all deals, and G2 measures which stage buyers find slowest, not how long it lasts. The chain from more reviewers to longer cycles is an inference, not a measurement.
A competing explanation is equally plausible for many deals. Buyers may be slower because budgets are tighter and payback expectations sharper, not because AI is special. G2's finance figures support that reading. If your AI deals close as fast as your non-AI deals of the same size, believe your own data over this post.
The sources also have interests. They include a review site, a revenue platform and an analyst firm that sells advice on buying groups. Their methods are stated, which is why they are cited here, but none is peer reviewed and the full Forrester and G2 reports sit behind client or registration walls.
How to measure enterprise deal velocity on AI deals
Enterprise deal velocity is usually reported as one number: open opportunities, times average deal value, times win rate, divided by cycle length in days. On AI deals that average hides where the time goes. Break it out by reviewer instead.
| Measure | How to collect it | What a bad reading looks like |
|---|---|---|
| Days from vendor selected to security sign-off | CRM stage timestamps, one new stage per reviewer | Longer than your whole evaluation stage |
| Reviewers named by week 2 | Count of the four reviewers identified and contacted | Fewer than three named |
| Days with no activity booked | Gap between scheduled activities per open deal | Any gap over 7 days, the threshold in Ebsta's data |
| Reversals after verbal approval | Deals lost after the champion said yes | A rising count quarter on quarter |
| AI deals against non-AI deals | Median cycle for the same deal-size band, split by AI scope | A gap that widens each quarter |
The last row matters most, and it is the one this post could not find in public data. Your CRM already holds it. Tag opportunities by whether AI features are in scope, then compare median cycle length inside the same deal-size band.
If you publish that comparison with the sample size and period stated, you will hold a figure no analyst firm has released. That is a better use of the data than a quarterly board slide.
Frequently asked questions
How long is the average enterprise sales cycle?
It depends heavily on deal size, and published averages vary by source. Gong's State of Revenue 2025, as summarised by SaaStr, puts $100,000 deals at about 70 days to close, though the summary gives no sample or period. Ebsta and Pavilion's 2024 benchmarks, covering 4.2 million opportunities, found first-half cycles 38% longer than in 2021. Measure your own median by deal-size band rather than trusting one figure.
How many people are in a B2B buying committee?
Forrester's State of Business Buying 2026 reports that 13 internal stakeholders and nine external participants influence the average B2B purchase, based on a survey of nearly 18,000 buyers. Purchases with generative AI features have buying groups about twice the size of those without, 14 members against 7. G2's 2025 survey found smaller core committees of 3 to 4, so the answer depends on whether you count deciders or influencers.
Why do AI software purchases take longer?
AI products touch sensitive data, produce unpredictable output and often carry usage-based pricing, so security, legal, privacy, AI governance and finance all want a review. G2's 2025 survey found about 8 in 10 buyers face stricter AI evaluation requirements from security, legal and compliance teams. No public source has timed the added days directly, so the longer cycle is inferred from more reviewers rather than measured.
How do you speed up a security review in enterprise sales?
Send the answers before the questionnaire arrives. Build an AI addendum to your security pack covering model providers, prompt and output storage, retention, tenant isolation and prompt injection testing. Get it to the named security reviewer within the first two weeks. G2's 2026 survey found IT security review is the biggest delay after vendor selection for 50% of enterprise buyers, so starting it late costs the most time.
Who is involved in AI procurement at large companies?
Typically the business owner and champion, IT, information security, legal or privacy counsel, an AI governance owner, procurement and finance. The IAPP's 2025 report found AI governance is owned by privacy teams at 22% of organisations, legal and compliance at 22% and IT at 17%. Forrester found procurement acts as a decision-maker in 53% of buying cycles, and G2 found finance involvement rose to 46%.
How do you measure enterprise deal velocity?
The standard formula multiplies open opportunities, average deal value and win rate, then divides by cycle length in days. For AI deals, add stage-level timing: days from vendor selection to security sign-off, days with no activity booked, and reversals after verbal approval. Ebsta and Pavilion found more than 7 days of inactivity cut win rates by 65%, which makes idle days a useful leading indicator.
Where to start this week
Pull your last ten AI deals that slipped or died after a verbal yes. For each one, write down which reviewer it was waiting on when it went quiet. If one reviewer appears in more than half, that reviewer's document is the first thing to build.
Then change your discovery script. Add one question to every first call: who, by name, will review security, data terms, AI governance and cost before signature? A champion who cannot answer has not yet bought anything, and the forecast should say so.
References
- Forrester, Barbara Winters, The State Of Business Buying: Risk-Averse Buyers Demand Proof, Not Promises, 21 January 2026. Used for buying group size and the genAI comparison.
- Digital Commerce 360, Mark Brohan, Forrester: B2B buying groups expand as they question AI, 22 January 2026. Used for procurement and C-suite roles and buyer views of larger groups.
- G2, New G2 Research: AI Is Reshaping How B2B Software Deals Are Won and Lost, 22 July 2026. Used for post-selection delays, finance involvement and CFO vetoes.
- G2 Learn, Sidharth Yadav, Why B2B Deals Stall After the AI Shortlist, 2 September 2026. Used for the 40% longest-stage figure, the 49% reversal figure and information security involvement.
- G2, New G2 Research: How AI is Redefining the Buyer Journey in 2025, 14 May 2025. Used for stricter AI evaluation requirements and shrinking core committees.
- Ebsta and Pavilion, B2B Sales Benchmarks 2024. Used for cycle change, slippage, inactivity and contacts engaged.
- IAPP and Credo AI, AI Governance Profession Report 2025, 16 April 2025. Used for AI governance adoption and ownership.
- SaaStr, Jason Lemkin, Gong: $100K deals take about 70 days to close, undated. Used for the days-to-close baseline from Gong's State of Revenue 2025.
The weakest point in this source base: no source here measures AI-specific cycle length in days, and the full Forrester and G2 reports are gated, so figures are taken from the firms' own published summaries. Figures are current as of 10 October 2026.
Related reading