Most GTM teams are still pouring money into the part of the funnel that already got easy. They optimize discovery, tune review-site presence, chase the shortlist, and celebrate when a logo lands in the top three of a buyer’s comparison. Then the deal goes quiet for six weeks, comes back with a security questionnaire and a new name on the thread from finance, and dies without anyone on the sales side ever knowing exactly why. The G2 2026 Buyer Behavior Report, based on a survey of 1,038 B2B software decision-makers fielded in June 2026, puts a number on what the field has felt for a year: evaluation is now the longest stage of the buying journey at 40%, having overtaken research at 36%. The friction moved to the one stage almost no revenue team has staffed, instrumented, or built a motion around..
AI moved the friction, it did not remove it
For two years the story was speed. Buyers found vendors faster, built shortlists faster, and got to a comparison faster than ever. That part is real. In the G2 data, 82% of buyers sourced software recommendations from an AI chatbot in the last 24 months, and half of those who used chatbots said the tools mattered most when narrowing and comparing options. Review sites now sit at the top of what shapes a shortlist at 38%, just ahead of AI chatbots at 37%. If your brand is absent from the AI answers and the review corpus, you are cut before a buyer ever contacts you.
Here is the trap. Everyone read that same data and doubled down on getting found. The G2 report is blunt about where that leaves you: AI made software easier to find and harder to buy. Getting onto the shortlist is now table stakes, and the shortlist only decides who gets to compete. The contest happens later, inside the buyer’s organization, in a stage that has quietly become the graveyard.
The mechanism is simple once you see it. Compress the front of the funnel and you shift the weight of the deal downstream.. Buyers arrive at evaluation faster, with more options, less internal alignment about why they even started, and a fresh set of AI-specific worries about cost and security. The Optifai analysis of stage-level CRM data across 939 B2B SaaS companies shows sales cycles have lengthened 22% since 2022, with the median now at 84 days and enterprise deals running 90 to 180 days or more. Two forces drive that stretch: larger buying committees, now averaging 6.8 stakeholders up from 5.4, and added security due diligence. Faster discovery and longer cycles are happening at the same time. That is the friction relocating to a room your reps are not in..
The two ways evaluation-stage deals actually die
When a deal stalls in evaluation, GTM teams tend to write it off as “no decision” or “lost momentum.” The G2 numbers name the two killers precisely, and both live downstream of the shortlist.
The first is the security review. IT security review is the single biggest source of delay after a vendor is selected, cited by 39% of buyers overall and 50% among enterprise buyers. Sit with that. Once you have won the evaluation on merit and been selected, the most common reason the deal then stalls is a review process that no one on the sales side owns. Budget approval follows at 32% and implementation planning at 25%, so the top three post-selection delays are all internal-buyer functions that a typical AE has no relationship with and no artifacts to serve. Optifai’s data reinforces the shape of it: security questionnaires, SOC 2 checks, and vendor risk assessments now add two to four weeks even in mid-market deals, and the negotiation-to-close stage alone accounts for 35 to 40% of total cycle time in enterprise.
The second killer is the CFO veto. This is the one that should change how you build your pipeline. In the G2 study, 49% of buyers say their CFO vetoed an already-approved software purchase in the last 12 months. Read that again: the purchase was approved, the buying team believed it was done, and finance reversed it. Finance involvement in software decisions jumped from 31% to 46% in a single year. The CFO is not a late signature anymore. The CFO is an active decision-reverser, and the reversal lands after your champion has already told you yes.
The AI category makes both killers worse. Among organizations with a dedicated token or LLM usage budget, the CFO-veto rate climbs from 49% to 54%, nearly double the 29% rate at organizations without one. And 80% of buyers say their organization now provides technical teams with exactly that kind of token budget. So the fastest-growing, most AI-native segment of the market is also the segment most likely to yank a signed-off deal because finance cannot predict what it will cost. The G2 report also found that concern about internal resistance to AI adoption grew from 16% to 29% in a single year, the largest single-year shift in the entire study. Your buyer’s biggest problem is increasingly their own colleagues.
Why sales never sees the deal coming apart
The reason this pattern persists is structural, not a matter of rep effort. The stages that AI compressed are the stages sales instruments obsessively: source of shortlist, demo booked, proposal sent, verbal yes. Those events happen in the CRM, on calls, in the seller’s field of view. The stages where deals now die happen in rooms the seller is not invited to. A security team runs a vendor risk assessment on its own schedule. A CFO questions a variable-cost line item in a budget meeting your champion attends alone. An internal skeptic raises an AI-governance objection in a Slack channel you will never see.
So the seller experiences the loss as a mystery. The deal was hot, then it went dark, then it came back dead. The forecast said commit. The instrumentation said healthy. The problem is that the instrumentation stops measuring at the exact point the real contest begins. This is the operator’s blind spot: you cannot manage a stage you cannot see, and you cannot coach a rep through a room they never entered.
Layer on the ROI problem and it compounds. In Gartner’s survey of 227 chief sales officers fielded in August and September 2025, 31% named difficulty proving the ROI of AI-driven tools as a top challenge for their 2026 sales objectives. If the people selling AI tools struggle to prove ROI, imagine the champion trying to defend that same purchase to a CFO who now reverses roughly half of approved deals. The seller cannot arm the champion because the seller does not have the proof, and the seller is not in the room where the proof is demanded. The deal dies of a defense that never got mounted.
The fix is a motion, not a document
The instinct is to treat this as a content gap: publish a security page, drop a SOC 2 badge, build a one-pager on ROI. That helps at the margin, and it is not the fix. Security readiness and champion enablement are a motion you run before the deal reaches evaluation, because by the time evaluation starts, the internal contest is already underway and you are not in it..
Here is what that motion looks like from the operator seat.
Start security readiness at the shortlist, not at the redline. The G2 data says security review is the top post-selection delay for 39% of buyers and 50% of enterprise buyers. So the moment a deal enters serious evaluation, you initiate the security track in parallel: send the completed questionnaire, the SOC 2 report, the data-flow diagram, and the subprocessor list before anyone asks. Optifai’s own tactics point the same direction. Deals with three or more contacts engaged close 2.4x faster than single-threaded deals, and mutual action plans that name the security-review window as a scheduled step cut cycle time meaningfully. Put the security review on the mutual action plan as a dated milestone with a named owner on both sides. If it is on the plan, it is a task. If it is not, it is an ambush.
Build the CFO defense before the CFO asks for it. With finance now reversing 49% of approved purchases and involvement up from 31% to 46% in a year, you cannot treat the CFO as a rubber stamp your champion collects at the end. Multi-thread to finance early and give your champion a defensible business case in the CFO’s own language: total cost of ownership, not seat price. This matters even more in AI deals, where 70% of buyers say the pace of innovation is pushing them toward shorter contracts and preference for outcome-based pricing more than doubled in a year, from 11% to 23%. Buyers who have already suffered a veto push for contract terms under 12 months at 40% versus 18% for those who have not. So arrive with pricing flexibility and a cost model your champion can defend, or watch a signed-off deal get reversed for a line item you refused to make predictable.
Own the internal-champion motion as a named stage. The G2 report is explicit that winning today means arming buyers with the proof and objection-handling they need to defend the decision internally. That is a job. Give it a stage in your pipeline, an owner, and exit criteria: champion identified, economic buyer mapped, security contact engaged, business case delivered in the buyer’s format, objection log built against the internal skeptics whose resistance jumped to 29%. Buyers who have been vetoed now expect positive ROI within six months of signing, so your business case has to survive that clock. If your CRM only tracks discovery, demo, proposal, and close, you are instrumenting the easy half of the deal and flying blind through the half that kills it.
What to build before the next quarter closes
The buyer has already decided how much authority to hand the machine, and it tells you where the human contest sits. Only 9% of buyers are comfortable letting AI agents execute purchases within approved guardrails, and just 2% would allow purchases without pre-approval, while 61% use or plan to use agents to research and shortlist. Agents will scout you. Humans will decide, defend, and sometimes veto you. The entire late-stage battle is human, internal, and political, and it is the part your current motion ignores.
For the second half of 2026, the move is to rebalance where GTM effort goes. The discovery and shortlist stages are largely automated and increasingly won or lost by presence in AI answers and review sites, so keep the lights on there and stop over-investing. Redirect the freed capacity into an evaluation-stage motion built for the two killers the data names: a security-readiness track that runs in parallel from the moment a deal gets serious, and a champion-enablement track that hands your buyer a CFO-proof business case and pricing they can defend against a finance reversal. Instrument both as named stages with owners and dated milestones so the rooms you were never invited to finally show up in your pipeline.
The teams that keep optimizing discovery will keep getting shortlisted and keep losing in a room they cannot see. The teams that build the internal-champion and security-readiness motion will win deals that were always going to reach evaluation anyway, because they will be the only vendor who showed up for the fight that actually decides it. Evaluation is the longest stage now. Staff it like you mean to win it.