A SaaS Management Policy People Follow Is a Speed Promise, Not a Rulebook

Saas Management Policy

A software purchase under $10,000 took an average of 55.8 days to get through procurement in the third quarter of 2026, according to Vertice’s cycle time data by contract value, drawn from more than $75 billion of spend it manages. A corporate card takes about five minutes. Every SaaS management policy is competing against that gap, and most of them lose.

That is the real reason policies get ignored. It is rarely that people haven’t read them. It is that the policy asks a team lead with a deadline in two weeks to wait two months for a $300 a month tool, and offers nothing in return. A policy people follow is less a rulebook than a promise about speed, written down, with service levels the organization actually keeps.

This guide covers why the usual policy fails, what to put in a short one that works, the three clauses most 2026 policies are missing, and how to tell whether yours is holding.

The policy already exists. That’s the problem.

Most mid-sized and large organizations have some form of software purchasing policy. The evidence says it isn’t changing behavior.

PagerDuty’s 2026 shadow AI survey, run by Wakefield Research with 1,250 office professionals at companies with at least $500 million in revenue, found that 86% believe their company has policies for AI tools. Yet among those who have used AI for work, 66% have used tools they believed were not allowed under company policy (full findings). At companies with 1,500 or more employees, that rose to 72%. And 39% said they would rather use AI without telling anyone than risk being told they can’t.

So the policy is known, and it is bypassed anyway. That pattern extends beyond AI. Gartner predicted that by 2027, 75% of employees will acquire, modify or create technology outside IT’s visibility, up from 41% in 2022 (Gartner cybersecurity predictions, March 2023). Zylo’s data shows where that spending lands: the average number of expensed applications rose from 125 to 138 while apps bought through normal channels fell from 229 to 206, and 77% of IT leaders discovered AI features or apps running without IT’s awareness (Zylo SaaS statistics).

The expense channel is growing because it is the fast lane nobody designed. A policy that does not offer a faster legitimate lane will keep feeding it.

Where the waiting actually happens

Before writing rules, look at what the current process costs the person asking. Vertice’s breakdown of procurement stage completion times for Q3 2026 puts the averages at:

Stage Average days (Q3 2026)
Intake 11.0
Legal approval 11.0
Customer negotiation 13.1
IT and security approval 7.3
Executive approval 5.7
Contracting 3.0

Two numbers stand out. Intake, which is simply getting the request into the system and routed to the right people, takes as long as legal review. And across Vertice’s overall cycle time data, new purchases rose from 39.3 days in Q2 2026 to 53.9 days in Q3, a 37% jump in one quarter.

Intake delay is pure policy design. It comes from unclear forms, unclear thresholds and requests that sit waiting for someone to decide who decides. It is also the cheapest delay to remove, because it needs no extra reviewers, only clearer rules about which lane a request belongs in.

The size of a purchase barely changes the wait at the low end. In Vertice’s contract value breakdown, purchases under $10,000 averaged 55.8 days and those from $10,000 to $50,000 averaged 70.3 days. A small team buying a small tool faces nearly the same process as a department buying a significant one. That is the single design flaw a good policy fixes first.

What goes in a one-page policy

The core policy should fit on a page or two. Procedures (the security questionnaire, the contract checklist, the renewal template) live in linked documents. The page itself should answer five questions a buyer actually has: which lane am I in, how fast will I hear back, who owns this once we have it, what happens at renewal, and what happens when I leave.

Here is a version you can adapt. The thresholds are starting points, not benchmarks; set them against your own spend profile.

Three lanes, each with a service level

Lane 1, the catalog (decision same business day). Any tool on the approved catalog can be provisioned by request to IT, with manager approval only. The catalog includes approved AI assistants, not only traditional apps. Most requests should land here, so the catalog has to be maintained, not published once and forgotten.

Lane 2, low risk (decision within 5 business days). New tools under a set annual value (for example $10,000) that touch no customer, employee or regulated data and do not process company content through an AI model. Requires a short security questionnaire and budget owner approval. If the organization misses the 5-day service level, the request is approved by default for a 90-day trial under standard terms.

Lane 3, full review (decision within 30 business days). Anything above the threshold, anything handling sensitive data, and any AI tool that processes company documents, code or customer content regardless of price. Full security, legal and finance review.

The default-approval rule in Lane 2 is the clause that does the most work, and the one legal and security teams will push back on hardest. It is worth defending. It turns the service level from an aspiration into an obligation on the reviewers, and it gives requesters a reason to use the front door. If a team knows that waiting 5 days guarantees an answer, far fewer will reach for a card.

The AI rule in Lane 3 matters because a $20 a month AI tool can carry the data exposure of a six-figure platform. PagerDuty’s data found 88% of respondents had shared work information with public AI tools, including 34% who had shared customer data (PagerDuty findings). Price is a poor proxy for risk once AI is involved.

Ownership

Every application above Lane 1 has a named business owner (a person, not a team) who is accountable for seat counts, renewal decisions and usage reviews. Ownership transfers on role change through the same workflow that handles access changes. Applications with no owner after 30 days go to a quarterly review list for retirement.

Renewals

Any renewal above a set annual value requires an affirmative decision from the owner at least 90 days before the renewal date, or before the cancellation notice deadline if that falls earlier. Auto-renewal without a recorded decision counts as a policy breach for the owner, not for the vendor. The owner brings usage data to the decision, which is the leverage a renewal negotiation depends on; our guide to negotiating SaaS contracts covers how to use it.

Leavers

Licenses are reclaimed through the identity provider as part of offboarding, within two business days of departure. Paid seats return to a pool before anyone buys new ones.

That is the whole policy. Anything longer, buyers won’t read.

Three clauses most policies still don’t have

The original version of this guide, written in early 2026, did not include any of the following. Each now covers a real cost exposure.

Default-on meters

Vendors are increasingly switching consumption billing on by default inside apps you have already approved. Microsoft enabled usage-based billing by default for new Microsoft 365 Copilot Business seats from October 19, 2026 (our breakdown). Atlassian starts billing extra usage on five meters, with extra usage on by default, from December 3, 2026 (details here). No purchase request fires, so a policy built only around new purchases never sees them.

The clause: any metered or AI component in an approved app requires the owner to set a spending cap or alert threshold, and new paid features stay disabled until the owner opts in. The metered portion is tracked as its own budget line. Zylo found 78% of IT leaders hit unexpected charges tied to AI features or consumption pricing (Zylo 2026 SaaS Management Index), so this is not an edge case.

Data portability for your own tooling

Productiv, a SaaS management platform, shut down on August 6, 2026, with about four days between announcement and sunset, according to a migration guide published by competitor Zluri. The guide notes that usage history was the one thing customers could not rebuild. In the same year CoreStack acquired BetterCloud (announcement) and Vertice acquired Vendr (CFOtech).

The clause: the systems that hold your SaaS inventory, usage data and contract records export to a store you control on a fixed schedule, at least quarterly. That applies to whatever platform runs your policy, because your renewal leverage lives in that history.

An amnesty path

When a new policy launches, the existing shadow portfolio does not disappear. People who bought tools on cards last year have every reason to stay quiet unless declaring them is safe. PagerDuty found 33% of people who would hide their AI use would do so to avoid scrutiny from managers or leadership (PagerDuty findings).

The clause: a fixed declaration window (60 to 90 days) during which any existing tool can be registered with no penalty, moved onto an owner and assessed through the normal lanes. After the window, unregistered tools found through discovery go straight to review. Pair the window with real discovery from expense, card and identity provider data, so the amnesty is a choice rather than the only source of truth. Our guide to shadow IT management covers the detection side.

Who owns what

The original version of this guide carried an eight-row responsibility matrix. In practice, five decisions cause most of the arguments:

Decision Accountable Must be consulted
Which lane a request is in SaaS governance lead Requester, security
Security and data approval Security Business owner
Budget approval Budget owner in finance Business owner
Renewal: keep, cut or renegotiate Business owner Finance, procurement
Metered caps and feature opt-ins Business owner Finance

The SaaS governance lead can be a fraction of one person’s role in a smaller company. What matters is that one named person can settle lane disputes in a day, because unresolved routing is where intake time goes. In larger organizations this role usually sits inside broader IT financial governance.

Enforce through the payment rails, not the document

A policy document enforces nothing. The controls that change behavior sit in systems people already use.

  • Card and expense rules. Flag recurring software charges on corporate cards and expense reports for review, and route them to the governance lead rather than straight to reimbursement. Given that expensed apps are the one category still growing in Zylo’s data, this is where most new shadow spend will show up.
  • Single sign-on as the gate. Apps that connect to corporate identity get provisioned and deprovisioned automatically. Make SSO a requirement for Lane 2 and Lane 3, and review OAuth grants in your Microsoft 365 or Google Workspace admin console, which is where AI assistants connect to company data.
  • The intake form itself. One form, a handful of questions, automatic routing by lane. If the form needs a training session, it is too long.
  • A published catalog. Requesters should see what is already approved before they ask for something new. Every new duplicate tool avoided is a rationalization project you don’t have to run later.

Then publish the service levels and report against them. Nothing builds compliance faster than a requester seeing that Lane 2 decisions really do come back in five days.

How to tell whether it’s working

Measure the policy on the promises it made, not only on the controls it imposed.

  • Request to decision time by lane, against the service level. If Lane 2 averages 12 days, the policy is broken regardless of what the document says.
  • Share of new SaaS spend that came through intake versus cards and expenses. This is the clearest single measure of whether people trust the front door.
  • Default approvals triggered. A rising count means reviewers are missing the service level; fix capacity, not the rule.
  • Owned spend share. Percentage of SaaS spend with a named owner and a known renewal date.
  • Metered variance. How far metered and AI spend ran from forecast each month, per owner.
  • Renewals decided on time. Every renewal that rolled over without a recorded decision is lost leverage.

Review the policy itself every six months. If exceptions exceed roughly one request in ten, the lanes or thresholds no longer match how the organization buys software, and the right response is to change the lanes, not to tighten enforcement.

A SaaS management policy works when following it is the fastest route to a tool. Everything else (discovery, SaaS spend management, software license management, renewal management) gets easier once the front door is quicker than the side door. For why the bill keeps rising even when the app count doesn’t, see our analysis of SaaS sprawl in 2026, and for the AI share finance can’t yet see, Finance Can’t See a Third of Your AI Spend.

ty247

Ty Sutherland is the Chief Editor at Kost Kompass. With 25 years of experience in enterprise strategy and financial management, Ty Sutherland is the driving force behind kostkompass.com. Specializing in helping Finance and Technology Managers optimize costs in servers, cloud, and SaaS, Ty combines technical acumen with financial discipline to deliver actionable insights for cost-effective solutions.

Recent Posts