Regrello, Agentforce, and Slack: scope the process

An architect exports every Flow in the org, highlights the list, and asks which ones Regrello will retire. The highlight is the mistake. Most of those Flows exist because a step must come out the same way every time. Assignment, a discount band, a field update from a rule you already trust. Those steps have an owner. Regrello automates complex processes, and Salesforce planned to integrate that automation with Agentforce and Slack. A complex process is not a synonym for every automation you have.

The line you can date is this one. "Salesforce completed its acquisition of Regrello on October 1, 2025." Anything that treats the close as an instruction to empty the Flow list has added a sentence the announcement did not contain. Keep the Flows that are doing a fixed job. Spend the quote on the process that still lives in inboxes, side chats, and a spreadsheet someone maintains because the path crosses too many teams for one Flow to be honest about it.

#A fixed step stays where it is

Take a discount. The band is a table. Quantity and product in, the figure the policy already sets out, under a policy finance already signed. That table is a Flow or an Apex method. It is tested. It does not become more correct because a process tool, or an agent, applies it in prose. The process may collect quantity and product. It may call the Flow. It may not keep a second copy of the band in a prompt or in a Slack message, because the second copy will be the one nobody updates.

Write that pattern down and reuse it. Entitlement milestones, tax, queue assignment from a matrix, a status that must follow from fields alone. If a new starter with the written rule would be wrong to produce a different output, the step is fixed, and it stays. The interesting process is the one where two careful people still have to talk: a supplier exception, a claim that touches three systems, an approval that waits on a fact nobody has typed yet.

Agentforce can start the process or answer a question about it. It is not a reason to delete the Flow the process calls. Slack can carry the message that asks a person to act. It is not the system of record for the discount. The record is still the record. If a Slack thread and a field disagree, you should already know which one finance will believe. Write that down before anyone builds the message.

#What belongs on the process list

Name processes, not platforms. "Vendor onboarding" is a candidate if you can say which teams touch it, which systems must be written, and where it stalls. "All operations" is not a candidate. For each process you keep, list the steps in order, and mark each step: Flow, person, or the process layer you are buying. A step with no mark is not ready to price.

Slack needs its own column. Which workspace. Which channel is allowed to start work, and which channel is only conversation. A message that creates a record is a write. A message that comments is not. People paste approval into chat because it is fast, and then nobody can reconstruct who agreed. If agreement matters, the agreement lands on a record, and the chat points at it. Say so in the brief. A partner who prices "Slack integration" without that split will build a notifier and call it a process, or a process that starts from any emoji.

Systems the process crosses, named by you. The contract store, the ERP, the org you already run. What each system owns. A process that writes the same fact to two systems will drift. Pick the owner. The other system receives a copy, or it does not. Copies need a direction and a failure you can see. Invisible sync is how the stall returns, one week later, as a dispute about which number was real.

Exceptions are the process. The happy path is usually short and already partly automated. The quote should dwell on the stall. Who is asked. What they need in the message. How long they have. What happens if they do not answer. A timeout that retries forever is not a design. A timeout that assigns a person, with the facts attached, is a design. Put the timeout in the scope. Leave it out and the pilot will look finished until the first absence.

#What cost means before anyone prices it

Cost is the named processes, the step marks, the systems, the Slack rules, the licences you already hold, and the hours of the approvers. This page has no day rate. Open your contract for Agentforce, Slack, and the platform products the process will call. If a product is missing, the quote may say it is required. It may not assume it is free.

Partner time is the build of the process layer for the steps you marked that way, the call into the Flows you are keeping, the Slack messages you allowed, and the test of one stall. Internal time is larger than teams expect. Approvers have to say the message is enough. An admin has to say which Flow remains the owner. A Slack owner has to say which channel may create work. If those people are not available, delay the fee. A fee that assumes their answers will be rewritten when they arrive, and the rewrite will be called scope creep.

Out of scope: retiring Flows that already produce a fixed output, a new ERP, a rewrite of every notification in the company, a claim that Agentforce now does the approvals because the process is integrated. Approvals stay with the person or the rule you named. Integration is not a delegation of authority.

Pilot one process. A second process is a second design, even when it shares a Slack workspace. You will learn from the first stall. Add the next process after that lesson, not in the same statement of work by default. Bundling five processes because they all sound operational is how none of them gets an owner.

#Migration services, for work that is not a record load

People will still search for a data migration quote under this name. The records the process touches may need a field, a status, or a link to an external id. That is configuration and a small map. It is not a licence to reload the org. If you also need to move historical cases or accounts, that is a data project with its own tools and its own rehearsal. Keep it off this line unless a specific object is required for the pilot process to see yesterday's work. Even then, limit the history to what the pilot reads.

What you are migrating, if anything, is the path that today lives in mail and chat. Write the path as it actually happens, including the ugly loop, before you draw the path you want. The ugly loop is the requirements. A clean diagram that skips the loop will be priced as the clean diagram, and the loop will reappear in the first month as an exception the tool cannot see.

Flow inventory is an input, not a backlog for deletion. Ask for a list of Flows the pilot process will call, and a list it must not modify. The second list is the safety rail. A partner who wants to rebuild those Flows inside the new layer should have to explain what the Flow gets wrong today. "It would be cleaner" is not a defect. Cleaner, with two owners, is a defect.

#The hiring test

People search for the best regrello Salesforce integration agency. Give the partner one process, the systems, and three Flows you believe are fixed. Ask them to mark every step. Ask which Flows they will call and which they will leave untouched. Ask which Slack messages may create a record. Ask what they will do when an approver does not answer. Ask which licence they need to see on the contract.

A pass marks the fixed steps as Flow and leaves them there. The stall has a person and a timeout. Slack is split into start-work and comment. The partner asks who the approver is, by role, not by a name that will be wrong next quarter. The partner refuses to retire the highlighted export as a programme of work.

A fail says every Flow should be replaced, or says it more softly by rebuilding them all. It prices Slack as a blanket integration. It lets a channel message stand as the approval. It calls itself the best agency. You can keep talking after the Flow list is restored. You should not sign while the list is still a deletion plan.

Second test: a step that changes money. The partner should show the existing rule and call it, or stop for a person. A new prose rule for the same money is a second policy. Two policies is how the exception becomes the norm.

#The first real stall

Run the pilot on live work, one process, with the approvers warned. Watch the first items that wait. If the message does not contain the fact the approver needs, fix the message before you add fields to the record for decoration. If the approver acts in chat and the record does not change, the write is in the wrong place. Move the write. Do not add a weekly reconciliation as the design. Reconciliation is the bill for a design that did not pick an owner.

Keep the Flow tests you already had. If a fixed step starts producing a different output after the process is connected, the process has grown a second writer. Disconnect that write. The connection can remain as a call. The Flow returns the result. The process stores it. That is the whole relationship.

How the process actually runs, including the loops a diagram skips, is a different question from automating it. Apromore and the event log is the page for discovering and simulating that path. Commission a log when you do not yet know where the work stalls. Commission this quote when you do, and you can name the stall.

Send one process and the three Flows you will not touch. Ask for the step marks before the fee. A partner who leaves those Flows alone, and who can describe the stall without a new policy, is ready to price a pilot. A partner who starts from the highlighted export is pricing a demolition.

#The other searches

The project quote is on regrello implementation cost. That page does not print a figure. Who bought the company is on what happened to regrello. The connector is on integrating regrello with Salesforce. Whether to move is on should I migrate from regrello. The customer fee is on regrello licence cost. Who to hire is on hiring for regrello. This page does not rank an agency.