The Questions to Ask Before You Commit to a Work OS
The demo will look flawless, because it was built to. These are the questions that find the wall you would otherwise hit three months after the contract is signed.
Choosing an all-in-one work platform is a decision you make rarely and live with for years, and the sales process is optimized to make it feel easier than it is. A demo is a curated path through a product's best moments; it tells you what the tool can do when someone who knows every shortcut is driving. Due diligence is the opposite discipline: deliberately walking off the curated path to find where the product actually ends. The questions below are the ones a vendor who has built a real platform will answer plainly, and the ones a vendor selling a thin one will deflect.
None of these are gotchas. They are the questions you will wish you had asked, phrased so that a straight answer tells you something and an evasive answer tells you more. Ask them before the contract, because after it your leverage is gone and the switching cost is what keeps you.
Questions about the data model
- Can an object in one part of the system natively reference an object in another - a task linked to a customer and a contract - or does connecting different kinds of work require an integration? This separates a shared model from products stitched together behind one login.
- What are the real limits: rows per table, items per view, automations per account, file sizes? Every platform has them. A vendor who claims none is either hiding them or has never had a customer big enough to find them.
- When two parts of the system show the same data, is it one record or two copies kept in sync? Copies that sync can drift; one record cannot.
- How do I get all my data out, in full, in a usable format, on my own initiative? The answer is a direct measure of how confident the vendor is that you will want to stay.
Questions about governance and security
- Is there one permission model across every kind of work, or does each module govern access separately? Separate models are separate gaps to leave open.
- Is there a single audit trail covering the whole platform, and can I export it? An audit that requires assembling logs from five subsystems is not an audit trail, it is a research project.
- When I remove a person, are they removed from everything at once? Fragmented offboarding is how access lingers after someone leaves.
- What are the specifics on data residency, encryption, backups, and third-party audits? Ask for documents, not adjectives. "Enterprise-grade" is a word; a current audit report is evidence.
Questions about migration and lock-in
The most expensive part of a work OS is rarely the subscription; it is the switch, both into the platform and, someday, potentially out of it. A vendor comfortable with their product will talk openly about both. One who only wants to discuss getting in is telling you they intend for getting out to be hard.
- What does a realistic migration look like for an organization my size, and who does the work - us, you, or a partner, and at what cost?
- What breaks or is not supported when we move our existing processes over? Every migration has friction; a vendor who claims none has not done one at your scale.
- If we left in two years, what would we be able to take, and what would we lose? The honesty of this answer predicts the honesty of the whole relationship.
- How does pricing change as we grow - more seats, more data, more automation? Get the growth curve, not just the starting number, because the starting number is the one designed to look good.
Questions about the parts they did not demo
Every product has a strong center and a weaker edge. The demo shows the center. Your job is to probe the edge, because that is where your real work will eventually land. Ask to see the feature that was mentioned in passing but never shown. Ask to build a small piece of your actual process, live, rather than watching the reference example. Ask what their own team does not use the product for, which is a question that produces a surprisingly candid answer more often than you would expect.
The point of pushing on the un-demoed parts is not to find a flaw and walk - every platform has soft edges, and a soft edge you know about is manageable. The point is to buy with your eyes open, so that the wall, if there is one, is a wall you chose to accept rather than one you discovered after migrating your company onto the wrong side of it.
How to weigh the answers
No product answers all of these perfectly, and a vendor who claims to is the one to distrust. What you are grading is not perfection but candor and fit: a vendor who names their limits, explains their trade-offs, and tells you what they are not good at is describing a real product you can plan around. A vendor who deflects, reframes, or answers a different question is telling you the limits exist and they would rather you find them yourself.
Atlas by WRX Stack is built to answer these plainly - one shared data model, one governance layer, a full data export on your own initiative, and honest limits rather than none. That is the standard the questions are designed to hold any all-in-one work platform to, including ours. Use them on every vendor you evaluate, and let the quality of the answers, not the polish of the demo, decide.