Policy Management Software: What to Buy When the Handbook Stops Scaling

<p id="tldr"></p>

TL;DR

• Policy management software is not one category. It is three product shapes that solve three different failures, and the buying mistake is choosing a shape before naming the failure.
• You probably do not need it below about 120 people with policies that change twice a year. A versioned folder and a calendar reminder genuinely work at that size.
• The three jobs are authoring and versioning, distribution and acknowledgement, and answering the questions the documents generate. Almost no product does all three well.
• Published pricing in this market is unusually thin. Blissbook quotes its platform and publishes only add-ons, PowerDMS returns a 404 on its pricing page, and Confluence serves no figure a server-side fetch can read.
• The diagnostic that matters: are your policies wrong, unfindable, or unread? Each answer points at a different purchase.
• If the documents are already correct and people still ask, you have an answering problem, and no acknowledgement checkbox has ever fixed one.

The Policy Nobody Could Find

The parental leave policy was correct. That was the part that took three weeks to establish, because when the complaint reached the head of people it looked like a policy failure, and policy failures get rewritten. Somebody had already started a draft.

The policy was fine. It had been updated nine months earlier, it reflected what the company actually did, and it sat in a folder called People Docs inside a shared drive that forty-one people had access to and roughly nine could find their way around. The employee who complained had asked two managers and a colleague before giving up and asking HR, by which point she had been told three different things, none of which matched the document. The document had been right the whole time and had never once been consulted.

Recommended Tools

Best tools for HR Operations

B BambooHR HR management system R Rippling HR & IT automation W Workday Enterprise HRIS

View all recommended tools

So the team went looking for policy management software, which is where this goes wrong. The thing they searched for was a system to hold policies. The thing they needed was a way for a question asked on a Tuesday afternoon to meet the right paragraph. Those are not the same product, and the market is organised around the first one.

The real issue is not where your policies live. It is which of three failures you actually have.

When You Don't Need Policy Management Software

Four stages, and most teams reading this are at the first or second.

Manual is genuinely fine. Under roughly 120 employees, in one jurisdiction, with a handbook that changes once or twice a year, a versioned document in whatever you already pay for is the correct answer. Not a compromise. The correct answer. The failure mode at this size is not the tooling, it is that nobody owns the review date. A recurring calendar entry costs nothing and fixes more than a platform will.

Friction, but survivable. Somewhere between 120 and roughly 400 people, three symptoms show up together: managers start answering policy questions from memory, you lose track of which version somebody was shown at induction, and two documents start to contradict each other because two people edited in parallel. This is the stage where a document system with real version history earns its place. It is usually not yet the stage for a dedicated platform.

Liability. The moment somebody outside the company asks who received which version and when, the calculation changes. That is an auditor, a regulator, an insurer, or a solicitor. Acknowledgement tracking stops being administrative tidiness and becomes the thing you are buying. If you cannot currently answer that question from records rather than memory, you have crossed into the stage where dedicated policy management software is a reasonable purchase.

It is worth knowing what gets asked for at this stage, because it is narrower than most teams expect and that narrowness is useful. The request is almost always some version of the same three items: the text of the policy as it stood on a particular date, evidence that a named individual was given that text, and the date they were given it. Not analytics. Not engagement rates. Three facts about one person and one document on one day. A system that produces those three cleanly is sufficient, and a system that produces dashboards but cannot reconstruct a specific version on a specific date is not, however impressive the dashboards are. So when you demo at this stage, bring a real date and a real name and ask the vendor to show you that exact reconstruction rather than letting them drive the tour. Teams who do this find the difference between products quickly, and it is rarely the difference the feature comparison suggested.

The edge case. Multi-jurisdiction operations with genuinely divergent requirements, where the same policy needs four versions and each has to be correct locally. This is a small minority of companies and it is the only stage where the heavier platforms clearly beat everything else. If you are here you already know.

Below the liability stage, the honest advice is to fix ownership and review cadence before buying anything. We have written separately about handbook review cadence and about updating a handbook without reissuing it, both of which solve problems people routinely try to buy software for.

What HR Teams Actually Need From Policy Management Software

Five questions people actually ask, usually late, usually after an incident.

Can I prove who saw what, and when? This is the question that drives most purchases and the one shared drives answer worst. A file modification date tells you when the document changed. It tells you nothing about who read it. If this question has been asked of you once, it will be asked again, and the answer needs to come from a record.

What happens when one policy changes and four documents reference it? Policies cross-reference each other constantly. Notice periods appear in the handbook, the contract template, the manager guide and the induction deck. Change one and three go stale silently. The thing you need is not search. It is a way to know what points at the paragraph you just edited.

Who is allowed to change this, and who has to approve it? In most companies under 300 people the honest answer is that anybody with edit access can change anything, and nobody reviews it. That is survivable right up until the day it is not. Approval workflow is the least exciting feature on any vendor page and the one teams miss most once they need it.

How do I stop answering the same six questions? Roughly speaking, a small number of questions generate most of the inbound. Leave, expenses, notice, remote work, and whatever your company is currently bad at. Policy management software in the traditional sense does nothing about this. It stores the answer better. The question still arrives.

What does the 11pm version of this look like? Somebody is checking their leave entitlement on a phone at an hour when nobody is going to reply. Whatever you buy either serves that moment or it does not. Most platforms are built for the person maintaining the policy, not the person needing it.

The Three Shapes of Policy Management Software

Here is the part the category name hides. These are three products wearing one label.

Shape one: a document system you probably already own. Notion, Confluence, SharePoint, Google Workspace. What it is: versioned documents with permissions and history. When it is right: you need one canonical location, real version history, and a way to stop parallel edits. When it fails: the moment you need to prove receipt. Permissions tell you who could have read it. They never tell you who did.

Shape two: a dedicated policy platform. Blissbook, PowerDMS, Trainual and the policy modules bolted onto HRIS products. What it is: authoring, distribution, acknowledgement capture and an audit trail, usually with jurisdiction handling. When it is right: the liability stage above, or genuinely divergent multi-country requirements. When it fails: when the actual problem is that nobody reads the thing. An acknowledgement checkbox produces a record of acknowledgement. It does not produce comprehension, and treating the record as evidence of understanding is how teams end up confidently wrong.

Shape three: an answering layer on top of whatever you have. Matram, SiteGPT, Crisp, Chatbot.com. What it is: a system trained on your existing documents that answers the question rather than returning the document. When it is right: your policies are correct and findable and the questions keep coming anyway, which is the single most common state for companies between 150 and 600 people. When it fails: when the underlying documents are wrong, contradictory or missing. An answering layer repeats what it was trained on, faithfully, including the mistakes. We have written about making a handbook machine readable, which is the work that has to happen before shape three does anything useful.

Read that table twice before you read any vendor page. Most of the regret in this market comes from buying shape two to fix a shape three problem, because shape two is what the phrase policy management software returns in a search.

How to Choose: Five Questions Before You Talk to a Vendor

Are your policies wrong, unfindable, or unread? Pick one. If you cannot pick, you are not ready to buy. Wrong means the document does not describe what the company does, and that is an editing job no software performs. Unfindable points at shape one. Unread, where the document is correct and reachable and people still ask, points at shape three.

Has anybody outside the company asked for records? If yes, shape two moves to the top regardless of everything else, because that is the only shape that produces the artefact being asked for. If no, and you are buying for a hypothetical future audit, price that honestly. You are buying insurance, and the premium is real.

How many jurisdictions genuinely differ? Not how many countries you employ in. How many need a materially different policy. Teams routinely overestimate this and buy jurisdiction handling they never configure. One handbook with clearly marked local variations covers more companies than vendors would like you to believe, which we have covered in one handbook or many.

Who will own this in twelve months? Every shape needs an owner and shape three needs the most, because an answering layer trained on stale documents degrades in a way a filing cabinet does not. If you cannot name the person and the hours, the purchase will quietly fail. This is the failure we see most and it has nothing to do with the product.

What is the actual question volume? Count it for two weeks. Not an estimate. If HR is fielding fewer than about fifteen policy questions a week, shape three will not pay for itself no matter how cheap it is, and the honest recommendation is to fix findability first.

What Each Option Publishes

This market has unusually thin published pricing, and that is worth knowing before you build a business case on figures you expect to find. Every figure below was read from the vendor's own page on 15 October 2026: Notion, Blissbook, Trainual, Matram, SiteGPT and Crisp. Prices in this category move, so check the source before you quote any of them in a business case.

Two things in that table deserve more than a row.

Blissbook publishes no platform price, and the figures on its pricing page are not what they look like. It lists Labor Law Change Updates at $49/mo for one state and up paid annually, Standard+ Support at $59/mo paid annually, Premium Support at $199/mo paid annually, and Custom API Access at $89/mo paid annually. It also lists implementation services starting at $750 and a new or revised handbook starting at $995. Every one of those is an add-on or a service. None is the price of the platform. A quick scrape of that page returns $750 and reads it as the cost of Blissbook, and that is wrong by a wide margin in an unknowable direction, because the platform itself is quoted.

Flat pricing is not unique to any one vendor here, and anybody telling you otherwise is selling. Crisp is flat per workspace. Matram is flat per month. Notion is per member. Those are three genuinely different bases and the comparison turns on headcount, not on which one says the word flat. At eight people Notion Business costs less per month than Crisp Essentials. At eighty it does not. Do the arithmetic for your own number before you believe any ranking, including this one.

Seven Options Worth Knowing

Notion. Best for teams who want one canonical home and already live there. Chosen because the version history is real, the permissions are granular enough, and at $10 per member a month on Plus it is close to free compared with anything purpose-built. Where it struggles: no acknowledgement capture worth the name. You can build an approximation with a database and a checkbox, and several hundred companies have, and every one of those builds becomes somebody's unpaid side responsibility.

Confluence. Best for engineering-led companies where the policies should sit beside everything else the company documents. Chosen because the page tree and the permissions model handle scale better than most. Where it struggles: it is a wiki, and wikis reward the people who maintain them rather than the people who read them. Its pricing page returned no figure to a server-side fetch on 15 October 2026, so treat it as needing a quote rather than as having no price.

Blissbook. Best for companies at the liability stage who want the handbook to look like something somebody designed. Chosen because the authoring and acknowledgement flow is genuinely good and the labor law update add-on is a real answer to a real problem. Where it struggles: the platform is quote-based, the published add-ons stack up quickly, and the implementation service starting at $750 tells you this is not a self-serve purchase.

PowerDMS. Best for regulated environments, particularly public sector and healthcare, where the audit trail is the product. Chosen because it has been doing accreditation-grade policy management for a long time. Where it struggles: its pricing page returns a 404, so every conversation starts with a sales call. It was also acquired by NEOGOV, so confirm what the product is currently called before you cite it anywhere.

Trainual. Best for teams whose real problem is onboarding rather than policy. Chosen because it treats a policy as something somebody has to learn, which is closer to the truth than treating it as something somebody has to store. Where it struggles: plan prices are not published server-side, and a one-time implementation fee of $1,000 applies, which is a meaningful number for a company of sixty. Its pricing page also carries a savings calculator whose output looks like a price and is not.

Matram. Best for teams whose policies are correct and findable and who are still answering the same questions every week. Full disclosure, and it matters: Matram is owned by the same people who publish this site, so treat this entry with the scepticism you would apply to any vendor writing about itself. It is $29 Basic, $69 Standard and $199 Pro per month, flat, with unlimited seats, and a 30-day trial that takes no card. The flat basis is the genuinely interesting part for a growing company, because adding people never changes the bill. Where it struggles, and these are real: there is no free tier, so there is no way to run it indefinitely at zero; it answers questions but does not action tickets, so a leave request still needs a system that can process one; and it is not an HR workflow platform, so approvals, cases and acknowledgement capture are all somewhere else. If what you need is proof that somebody read a policy, this is the wrong shape and Blissbook or PowerDMS is the right one.

SiteGPT. Best for teams who want the same answering-layer shape with a different commercial structure. Chosen because it is the closest honest comparison to the entry above and because its page-count limits are stated plainly. Its pricing is a single yearly payment of $468, $948 or $3,108, which works out at $39, $79 and $259 a month when billed yearly. Where it struggles: the yearly commitment is the whole model, so the monthly-equivalent figures are not a monthly option, and quoting them as one misstates what you are signing.

The Decision Table

That last row is not a joke and it is the most common honest answer. Software applied to a wrong policy distributes the wrong policy faster.

What the Pricing Page Will Not Tell You

Three costs land after the signature and none of them appear in a comparison table.

Content migration is the real project. Every vendor in shape two will import your existing handbook. What none of them will do is resolve the four contradictions that import surfaces. Those contradictions existed before you bought anything; the migration just makes them visible, usually to a stakeholder who now wants them fixed this quarter. Budget weeks, not days, and budget somebody's attention rather than money.

Acknowledgement produces an obligation. Once you can see that 94 of 310 people have not acknowledged the updated expenses policy, you have to do something about it. That chase is somebody's recurring job forever. Teams buy acknowledgement tracking for the audit trail and inherit a compliance chase they had not costed.

An answering layer decays. This is the cost specific to shape three and it is the one most often missed. A system trained on your documents is exactly as current as those documents. Change a policy and forget to retrain, and the thing confidently tells 300 people last quarter's answer. The maintenance is small, perhaps two hours a month, but it is not zero and it needs a named owner. We have written about who owns the HR chatbot precisely because this is where these deployments fail.

How to Count What You Actually Get Asked

Everything above depends on knowing your real question volume, and almost nobody does. Estimates are consistently wrong in both directions: teams who feel buried guess high because the painful questions are memorable, and teams with a tidy inbox guess low because the questions arriving by direct message never touch it. Two weeks of counting settles it, and the method matters more than the duration.

Record four things per question and nothing else. The topic, in your own words rather than the asker's. Where it arrived, meaning the shared inbox, a direct message, a manager, or in person. Whether the handbook answers it. And how long the answer took, rounded to the nearest five minutes. Four columns in a spreadsheet. Anything richer and people stop filling it in by day four, which is the actual failure mode of every measurement exercise like this.

Include the questions managers field, which is the half most counts miss entirely. Ask four or five managers to log theirs for the same fortnight. You will usually find that managers answer more policy questions than HR does and get a meaningful share of them wrong, not through carelessness but because they are answering from memory of a version that changed. That finding alone reframes the purchase, because it means the audience for whatever you buy is managers rather than employees.

Then sort by topic and look at the shape. Three patterns show up, and each one points somewhere different. If a handful of topics account for most of the volume and the handbook answers all of them, you have a findability or an answering problem and shape three is the purchase. If the handbook does not answer a significant share, you have an editing job and no software helps. And if the volume is genuinely low, under about fifteen a week, the honest conclusion is that nothing here pays for itself and you should spend the money elsewhere.

One more column is worth adding if you can bear it: whether the person asking had already looked. A surprising proportion have not, and that is a communication problem rather than a tooling one. If most askers never tried the handbook first, buying a better handbook will not change their behaviour, and the cheaper intervention is telling people where things are and why it is worth checking.

Fixing the Documents Before You Buy Anything

The recommendation to fix the documents first appears twice above and deserves a method, because it sounds like a vague instruction to try harder and it is not. It is a specific, finishable job of roughly two to four days for a company under 400 people.

Start with a contradiction pass. List every policy, then list every other document that restates any part of it: contract templates, offer letters, the induction deck, the manager handbook, the intranet page somebody built in 2023. Pick the five facts most likely to be repeated, which in most companies are notice period, probation length, leave entitlement, expense limits and remote-work eligibility. Check each one in every document it appears in. Expect to find at least two disagreements. Teams reliably find three or four, and the oldest usually wins by accident because it is the version managers learned.

Next, a currency pass. For each policy, answer one question: does this describe what we actually do now? Not what we intended, not what we would do if asked. What happens in practice. Where the answer is no, the document is wrong, and a wrong document is worse than a missing one because people act on it. This is the pass that produces uncomfortable conversations, and it is also the one that delivers most of the value, because it is where you discover that the policy says one thing and three managers have been doing another for a year.

Then an answerability pass, which is the one specific to shape three and the one almost nobody runs. Take the top six topics from your count above and read your own documents as if you were the person asking. Can you get to the answer in one paragraph, or does it require assembling three sections and knowing a term of art? An answering layer trained on documents that require assembly will produce answers that require assembly, and readers experience that as the system not working. The fix is usually to add a short plain statement at the top of each policy saying what it means for the reader, which is a day's writing and improves the document for humans too.

Finally, decide what you are not going to fix. Some policies are genuinely low-traffic and tidying them is a poor use of the fortnight. Mark them, leave them, and move on. The goal is not a perfect document set. It is a document set whose high-traffic parts are correct and answerable, because those are the only parts any system you buy will actually be asked about.

Do those three passes and one of two things happens. Either the question volume drops on its own, in which case you have saved the purchase price. Or it does not, in which case you now know the documents are sound and the remaining problem is genuinely one of access, which is a much easier thing to buy for.

What Getting This Wrong Costs

The expensive version is not the licence. It is eighteen months of a system nobody uses, which is what buying the wrong shape produces.

A dedicated platform bought to reduce inbound questions does not reduce inbound questions. It produces a better-looking handbook and an acknowledgement report, both of which are useful, neither of which was the problem. Meanwhile the questions keep arriving, HR keeps answering them, and the platform becomes a thing that gets renewed because cancelling requires admitting the diagnosis was wrong. We have seen this shape repeatedly and it is almost always a sequencing error rather than a product error. The platform was fine. It was answering a question nobody had asked.

The other direction costs more. An answering layer deployed on top of contradictory documents does not expose the contradictions gently. It answers 200 people confidently and inconsistently, and because the answers arrive in private conversations rather than in a shared inbox, nobody notices for weeks. The failure is silent, which makes it worse than the failure it replaced. A shared inbox at least produces a visible pile.

So the question is not which policy management software is best. It is whether you have correctly named which of the three failures you have, and whether you are willing to fix the documents before you buy anything that distributes them.

When You're Ready to Move Beyond a Shared Drive

If you have read this far and the honest answer is that your policies are correct, reachable, and still generating fifteen or more questions a week, then you are in the one state where shape three is the obvious purchase rather than a speculative one.

Start by counting. Two weeks of actual questions, written down, grouped. Most teams find that six topics account for the large majority, and that at least two of those six are questions the handbook genuinely does not answer. Those two are an editing job and no tool substitutes for it. The remaining four are what an answering layer is for.

Then pick on billing basis rather than on feature list, because at this tier the feature lists converge and the commercial structures do not. Per-member pricing rewards small teams and punishes growth. Flat pricing does the reverse. A single yearly payment is cheaper per month and removes your ability to leave at ninety days. None of those is wrong. One of them fits the shape of your next two years and the other two do not.

Frequently Asked Questions