Nobody knows where a request currently is
Answering "what happened to mine" means searching an inbox, and the answer is often that it is sitting unread with someone on leave.
Requests that need a second participant, a decision and a record end up in email — where there is no route, no escalation and no way to report on any of it. Twopir Consulting models those processes in FormAssembly's Workflow Builder: approval steps, conditional routing and connectors placed at the step where the data becomes true. More than one person means more than one form.
Trusted by 500+ organizations — including healthcare, higher education, nonprofit and financial services teams whose approvals currently happen in an inbox.










Built for Regulated Data Collection
Email is an excellent messaging system and a poor workflow engine. It has no concept of a pending state, an owner, a deadline or an audit trail — so every one of those has to be supplied by a person remembering. At volume, somebody always stops remembering.
Answering "what happened to mine" means searching an inbox, and the answer is often that it is sitting unread with someone on leave.
An approver who does not respond produces silence, not a reminder. The request waits until the requester chases it, which is a queue managed entirely by persistence.
The decision exists in a reply that says "fine by me". For anything with a financial or regulatory consequence, that is not an audit trail.
Everyone knows requests over a threshold need a second signature. Whether that actually happened on any given request depends on whether the person remembered.
The second participant re-enters what the requester already provided, because there is no shared record between the two steps — only a forwarded message.
You can count requests and count outcomes. You cannot say where time is lost, which approver is the bottleneck, or whether last quarter was better — because the process was never data.
FormAssembly Workflow is a separate part of the product from the form builder. A workflow is assembled from steps: forms completed by different participants, approval steps where a named decision-maker approves or denies, conditional steps that route down different paths based on submitted values, steps that jump the process elsewhere, and connectors that run at a specific point rather than at the end. The distinguishing property is that a workflow can pause — for hours or for days — and each step can belong to a different person, inside or outside your organization.
Where this page stops. If the process is one person filling in one thing, even a long thing, it is FormAssembly form automation and it should not be built as a workflow. How each individual form looks and what it asks is FormAssembly customization services. The mapping and matching logic behind a record write is FormAssembly Salesforce integration. Where an approval's evidence trail exists to satisfy a regulator, pair this with FormAssembly security and compliance consulting.
What Twopir does, and what the product does. The Workflow Builder, approval steps, conditional rulesets, go-to-step routing and workflow-native connectors are FormAssembly capabilities. What we contribute is the process design: what the approval policy actually is once it is written down rather than assumed, where thresholds sit, what happens on silence, what the denied path does, and which step each connector belongs to. As a Salesforce Partner and HubSpot Partner we also build the reporting that makes cycle time visible afterwards. Product documentation is the FormAssembly Resource Center.
Every step you add is a policy decision you now have to make explicitly. That is the value of modelling the process — not the automation, but being forced to say what the rule actually is.
| Step type | What it does | The decision it forces |
|---|---|---|
| Form step | Collects information from a named participant at this point in the process. | What this participant may see of what has already been submitted. |
| Approval step | A decision-maker approves or denies, and the process continues down the matching path. | Who decides, whether one approver is enough, and what the denied path does. |
| Conditional step | Routes down one or more paths based on rules evaluated against submitted values. | What the policy thresholds actually are, in numbers rather than convention. |
| Go-to step | Sends the process to another point, rather than straight on. | Whether a denial means rework and resubmission, or termination. |
| Connector at a step | Creates or updates records at this point instead of at the end of the process. | When the data becomes true enough to be worth writing to the CRM. |
Assembling steps is the straightforward part. Deciding what the process actually is, and what happens when it goes wrong, is where the engagement earns its cost.
Most organizations discover during this step that the policy everyone follows has never actually been agreed in the same terms.
The rule enforced by the system rather than remembered by a person — which is also what makes a policy change a configuration change.
Each step belongs to someone, and what they can see is a decision — particularly when a participant is outside your organization.
Approval processes rarely fail on the decision. They fail on the days of silence before it, and that is the part nobody configures.
Records written when the data becomes true, so the CRM knows about work in progress rather than only about finished work.
The reason to model a process as data. Once it is steps rather than email, the process itself becomes reportable.
Durations describe typical engagements for one process. The variable that moves the timeline is almost never the build — it is how long it takes to get the approval policy agreed between the people who own it.
How the process runs today, every hand it passes through, and a cycle-time baseline taken before anything changes. One to two weeks.
Thresholds, approval levels, delegation, escalation timing and what a denial means — written down and signed off by the people who own the policy. One to three weeks.
Steps, forms, approval and conditional routing assembled, with connectors placed at the step where the data becomes true. Two to four weeks.
Every path walked deliberately: approved, denied, escalated, delegated, abandoned mid-step and returned for rework — not just the one where everyone says yes. One week.
Go-live with reporting in place and the stage-one baseline as the comparison, so the result is measured against your own process. One week, then optional retainer.
Each of these has more than one participant, a decision with consequences, and a record somebody will eventually ask for. That combination is the signal that a process should be a workflow.
Scholarship and programme applications needing a referee's input and a committee decision. Two external participants, a threshold, and a record that must survive an appeal.
Purchase requests, expense exceptions and discount approvals, where the threshold bands are exactly the thing currently enforced by memory rather than by the system.
New starter setup, access requests and role changes — several departments each holding one piece, and an access decision that a security review will later ask about.
Grant applications and beneficiary assessments with a review panel. The denied path matters as much as the approved one, because applicants are told why.
We’ve started to use FormAssembly to assess impact of the program by collecting self-reported measurements on a set of capabilities before people undertake the program and after completion and comparing them. That’s been incredibly useful and quite groundbreaking.
We have not published a FormAssembly-scoped case study, and we will not present one that does not exist. The closest published evidence of how we make a process reportable end to end is this integration engagement.
We help growing and mid-market companies solve complex CRM, integration and business system challenges, and we work with enterprise organizations on the same problems at larger scale.
Most organizations find during this step that two people describe the same approval rule differently. Finding that out before the build is considerably cheaper than after.
Most workflows are built as if everything gets approved. Rework, resubmission and a reason the requester can read are the paths that determine whether people trust the system.
Escalation, reminders and delegation during absence are configured, not hoped for. Silence is where approval processes actually lose their days.
As a Salesforce Partner and HubSpot Partner we build the CRM side too, so leadership sees work in progress rather than only completed requests.
Cycle time measured before go-live, so the improvement is a comparison against your own process rather than a generic claim about automation.
A multi-page form is still one person, in one sitting, filling in one thing. A Workflow is a process built in FormAssembly's Workflow Builder from separate steps — forms completed by different participants, approval steps where someone decides, conditional steps that route down different paths, and connectors that run at a particular point rather than at the end. The practical difference is time and ownership: a workflow can pause for days waiting for an approver, and each step can belong to a different person, which a multi-page form cannot do.
An approval step names the form response requiring approval, who decides, and what happens next. The decision produces two paths — approved and denied — and both are designed rather than one being an afterthought. Around that, the choices that matter are how many levels of approval a request needs, whether a threshold routes it to a more senior approver, what happens when the approver does not respond, and whether approval by any one of several people is sufficient or all of them must agree. Those are policy decisions we get written down before anything is built.
Yes, and this is usually where the real value is. A conditional step routes the workflow down one or more paths based on rules you define against the submitted values, so a request can skip approval entirely under a threshold, go to a team lead in a middle band, and escalate above it. Modelling the policy this way means the rule is enforced rather than remembered — and when the policy changes you change the ruleset instead of retraining everyone.
Whatever you design, which is the point — the default is that the request waits indefinitely and nobody notices. Escalation and reminder design is a distinct piece of work: how long a step may sit before it is chased, who is reminded, when it escalates past the original approver, and whether a delegate can act during an absence. Most approval processes that fail in practice do not fail on the decision, they fail on the silence before it.
At the step where the data becomes true, not at the end of the process. Workflow-native connectors let a record be created or updated at a specific step, so a request can appear in Salesforce when it is raised, be updated when it is approved, and trigger downstream work only on the approved path. Putting every connector at the end is the common shortcut, and it means the CRM knows nothing about a request until the process has already finished — which removes any possibility of reporting on work in progress.
Yes — a step can be completed by an external participant such as a supervisor, referee, guarantor, vendor or partner. The design questions are about access rather than routing: whether that participant needs to authenticate before opening their step, what they are allowed to see of what has already been submitted, and how long their link remains valid. Those are decisions to make deliberately, particularly when the earlier steps contain data the external participant has no business reading.
By reporting on the process, not just the submissions. Once a process is modelled as workflow steps rather than email, you can answer how long each step takes, where requests stall, how many are denied and why, and which approvers are the bottleneck. That was almost certainly unanswerable before, because the decision was happening in an inbox. We treat the cycle-time baseline as part of the delivery, so the improvement is measured against your own process rather than asserted.
How long from request to decision, and how much of that is waiting rather than working? If nobody can answer without searching an inbox, that gap is the whole argument for modelling it as a workflow — and the baseline you will measure the improvement against.
Or contact the team — we design the denied path, the escalation and the silence