The form is the easy part. The routing rules are where intake projects get abandoned.

What makes a Wrike request form different from a form tool?

The submission becomes a task in Wrike, with answers mapped to custom fields, and the task inherits the project’s workflow and automations.

Three consequences people miss:

• Answers are filterable. A “budget range” field means you can report on budget without reading tasks.
• The requester can be added as a collaborator, so status updates go out automatically via the task’s own notifications.
• Conditional sections let you hide questions that do not apply, which is the only real defence against form fatigue.

If you just need a form and no task, use any form tool. The value here is the task.

How do you build the marketing intake form?

Fields first, then routing. Build in this order:

1. Create custom fields on the folder: Request type (dropdown), Priority (dropdown), Budget range (dropdown), Launch date (date), Requester team (dropdown). Five fields, no more.
2. Create the form from the folder and attach those fields as questions. Mark only Request type and Launch date as required.
3. Add conditional logic on Request type: if “Design”, show a link field for brand assets; if “Paid campaign”, show Budget range. Conditional questions are third and below — never first.
4. Write a 2-sentence intro at the top of the form: what happens after submission and the expected first response time. That line reduces duplicate submissions more than any field.
5. Test with five real submissions from the team before launch, then delete the test tasks.

The intro sentence is the most valuable field on the form and it is not a field. Without a stated response time, people submit again in three days.

How do you make the routing automatic?

One automation rule per request type, on the destination folder.

A workable set for three types:

The order matters: Wrike runs rules in the order they appear, so put the escalation rule above the general assignment rule or both fire and the second overwrites the first.

That ordering bug bit me once. A $120k campaign went to the design queue because the general rule ran last.

What about AI triage on submissions?

Wrike’s AI can classify incoming requests and suggest the destination, which is genuinely useful when your form allows free-text descriptions.

It is less useful when your form has strong dropdowns, because the dropdowns already did the classification. The teams that get value from AI triage are the ones collecting messy prose — support tickets, inbound leads, free-form requests.

Availability: AI features require Business ($28.75/user/month) and above; forms themselves are available on most paid plans including Team at $9.80. If your only goal is structured intake, Team plus good fields beats Business plus AI.

Where request forms fall down

Four recurring problems:

• Too many questions. Eight fields reduces completion by roughly half in practice. Six is the ceiling; five is better.
• No default assignee. A submission with no rule falls into limbo. Always have a catch-all rule that assigns to a named triage person.
• Requesters who cannot see status. If you do not add them as a collaborator, they will email to ask. One checkbox prevents most of that.
• Duplicate submissions. The form does not detect them. Add an instruction to search before submitting, and accept that some duplicates will happen.

None of these are technically hard. All of them are the reason an intake form works for six weeks and then quietly stops being used.

Who should build this

Build it if you receive more than 20 requests a month through email or chat, you have at least three destination teams, and someone will maintain the rules. Marketing, IT help desks and internal creative teams are the strong fits.

Do not build it if you have one team and 10 requests a month — a shared inbox with labels is cheaper and nobody has to learn anything. And do not build it on the Free plan; forms and automations need a paid tier, and the Free plan’s 5-user cap will not cover a requester population anyway.

Quick Summary

• Wrike request forms create a task per submission with answers mapped to custom fields; five fields is the practical maximum.
• Add a 2-sentence intro stating what happens next and the response time — it cuts duplicate submissions.
• One automation rule per request type; put escalation rules above general assignment or the last rule wins.
• Add requesters as collaborators so status updates go out without anyone emailing to ask.
• Forms work on Team ($9.80/user); AI triage needs Business ($28.75/user) and mostly helps with messy free-text intake.

Risk prediction is the only feature that changes how you manage. The rest is convenience.

What does Wrike’s AI actually predict?

Wrike is work management for larger teams: projects, tasks, subtasks and folders, with Gantt charts, workload views and custom request forms.

The risk indicator looks at tasks that are overdue, tasks with no assignee, tasks whose due dates moved repeatedly, and dependency chains where an unstarted task blocks a started one. It surfaces a flag on projects and tasks that match.

It does not forecast hours or model scenarios. “What if we push this two weeks” is still a manual exercise with the Gantt.

Why do risk flags fire on the wrong projects?

Because the flags are computed from your own task data, and most Wrike accounts have tasks in the wrong state.

Common causes, from accounts I have looked at:

• Tasks left in “In Progress” after work finished. Every one of those is a false overdue signal.
• Placeholder dates. A team that enters “end of quarter” as a real date trains the model to expect impossible deadlines.
• No dependencies recorded. If tasks do not declare blockers, the chain analysis has nothing to analyse.
• Owner of last resort. One person assigned to 80 open tasks makes the workload suggestion meaningless.

The AI is not wrong; it is reading the data you gave it. Cleanup of these four things is the whole prerequisite.

How do you calibrate the risk feature in week one?

Seven days, four steps.

1. Close every finished task. Archive completed items. This takes an afternoon and is the highest-value cleanup.
2. Set a realistic default in the project settings so new tasks do not inherit placeholder dates.
3. Add dependencies to your three biggest projects. Not all of them — the three that matter, where you would notice a bad prediction.
4. Compare the flags to your own judgement for a week. Note every flag you disagree with and ask why. Nine times out of ten the task data explains it.

After that week you will know whether the feature is useful for you, and you will have improved your data regardless.

How good are the assignment suggestions?

Mixed, and the reason is structural. Suggestions weight current workload heavily, so they recommend the person with fewer open tasks over the person with the right skills. For routine work that is fine. For specialised work it is actively wrong — you do not want your only Kubernetes person bypassed because they are busy.

Practical rule: use the suggestion as a prompt to ask “why not this person”, not as an assignment. The workload view in Wrike is more useful than the AI suggestion, and it is deterministic, so you can argue with it.

What does Wrike cost, and which tier has AI?

Minimum seat counts apply on Business — usually five or more users. A team of six moving from Team to Business pays about $1,140 a month to get AI, which is a lot to pay for risk flags on data you still have to clean.

Where Work Intelligence disappoints

The editing features — draft, rewrite, summarise a task description — are generic. They work, they save a minute, and they make every task description sound the same. Teams notice within a week and stop using the rewrite button.

The risk flag is better but blunt: it tells you a project is at risk without telling you which decision caused it. You still open the Gantt and trace backwards. That is 10 minutes of work the AI saved you from discovering, not from doing.

One genuine strength: request forms with AI triage will route incoming marketing briefs to the right team without a human reading them. That is the feature I have seen actually save hours.

Who should use it

Use it if you are on Business already for the reporting, you run 10+ concurrent projects, and one person owns data hygiene. At that scale, catching one slipping project a month pays for the licence.

Do not buy Wrike Business for AI alone. The prerequisite cleanup is weeks of work, and the payoff is a flag you could get from a saved filter on overdue tasks. Small teams under 10 should stay on Team or Free and use filters; they will get 80% of the value for none of the cost.

Quick Summary

• Wrike AI = project risk prediction, workload-based assignment suggestions, and editor drafting. It is on Business ($28.75/user/month) and above.
• Risk flags are computed from your task data: stale “in progress” tasks and placeholder dates create false alarms.
• Close finished tasks and add dependencies to your top three projects before trusting any prediction.
• Assignment suggestions favour free capacity over skills — treat them as questions, not answers.
• Best real-world payoff is request-form triage, not the editor features.