Salesforce AIforce: How AI becomes a colleague at NY.Club
The team is scheduled. The bartenders know when they are working, and so do the runners. Everything is in Slack. We should be done. Should be.
Because now those shifts still need to be created in Salesforce. The same people, the same events, the same information. Just entered again in another system.
Tasks like these are part of running a club. Alongside the music, the guests and the late nights, there is staff scheduling, paperwork, payments and the monthly close. Since we took over NY.Club, I have built much of this into Salesforce.
AI is now changing how I work with that system. Two approaches in particular have caught my attention:
Coworker within Salesforce, and Salesforce through Claude — “Claudeforce”.
What started as experimentation now gets real work done. I estimate it already saves me six to ten hours a month. At the same time, it gives me insights into our business that were previously buried in the data and difficult to put to use.
A new colleague needs context first
After activating Coworker, I connected Slack and started by explaining what we use our Salesforce instance for: the processes it supports and how our club is organised. It felt surprisingly familiar. I would start by explaining how the business works to a new colleague, too.
Then I wanted to see how far Coworker could get with that context. Staff scheduling was a good test. We use Slack to ask our team when they are available. We then decide who will work at each event and share the schedule in the relevant channels, identifying team members through @mentions.
I had already built an Agentforce agent to create shifts in Salesforce. It could do the job if I told it who was scheduled. Until then, I still had to connect the finished schedule in Slack with that agent myself.
So I asked Coworker to check Slack and create the shifts for that day’s event in Salesforce. And sure enough, Coworker created them. Of course I had hoped it would work. What really impressed me was that I did not have to supply the names or explain how to get from Slack to Salesforce.
Coworker found the schedules in the channels. It recognised that the people mentioned also existed as contacts in Salesforce and could be matched using the Slack ID stored there. And it used the existing agent to create the shifts.
Not every search succeeded immediately, either. I could see Coworker finding nothing at first, broadening its search, checking individual channels and examining Salesforce metadata. It worked its way towards the solution.
The impact extends all the way to payment
Of course, this saves me time. But for me, the bigger change is the continuity of the process. We collect availability in Slack. That is also where the schedule takes shape. It becomes specific shifts linked to specific events in Salesforce. Later, the payment can be assigned to that exact shift.
Until now, I was the interface between Slack and Salesforce myself. Read the schedule, match the team members, create the shifts. That is exactly the part Coworker now takes over. It brings together information and capabilities that were already there and works out what the task requires. That is how I see its role as an orchestrator.
With Claudeforce, experimentation became part of the monthly close
I tried Claudeforce as soon as it was introduced. Initially, I used it for smaller tasks: retrieving information, updating data and finding out what worked. It became particularly useful for larger tasks spread across several systems.
We keep our cashbook in Salesforce. The supporting documents — invoices, receipts and other records — are stored in Google Drive. For the monthly close, these need to be brought together and prepared for DATEV. I can give Claude a single, connected assignment:
Go through the documents, compare them with the cashbook entries in Salesforce, link existing entries and create any that are missing. Where a payment is still outstanding, prepare the necessary cash withdrawal. The result is a folder of checked documents ready to pass on to DATEV. At the same time, the documents are linked to the entries in Salesforce and their associated accounts.
Outstanding payments are linked to upcoming events. This also lets the duty manager know what needs to be paid out that night.
The same approach works for expenses I have covered personally. A credit card statement might include Motel One hotel stays or Lufthansa flights for DJs. Claude can identify these items, match them with the relevant records in Salesforce and schedule reimbursement as an outstanding payment. This substantially reduces the effort involved in my monthly close. Above all, it leaves fewer loose ends between the supporting document, the entry and the payment.
My real highlight: new ways to look at the same data
Things got really interesting for me with artifacts. Suddenly, I could have views built that I had wanted for ages, but somehow something else had always been more important than actually building them. These are interactive views and small applications that Claude creates for my specific questions. One example is our year-on-year comparison. For a club, simply comparing the same calendar date in two different years often is not enough. I want to compare Fridays with Fridays, Saturdays with Saturdays, and whole weekends with one another.
With just a few instructions, I was able to have a view built using our Salesforce data: bar revenue, admission, cloakroom revenue, guest numbers, costs and trends over the selected period. This lets me understand how comparable club nights are developing and which areas contribute to a change.
Previously, I would have had to build complex reports and dashboards or use an additional analytics solution. For this particular analysis, I can now get to the view I need much faster.
The information was already there. But getting from “We have the data” to “I can use it to make a decision” often took considerable work.
That journey has become much shorter for me.
Where does the money come from — and where does it go?
Another artifact visualises our income and expenses as a flow of money. On one side, I can see where revenue comes from, such as the bar, admission or the cloakroom. On the other, I can see how expenditure is distributed across staffing, artists, drinks purchasing and other cost categories. This helps me understand a month’s financial picture at a glance. How do income and expenses compare? Which cost categories carry the most weight? What is left at the end?
Salesforce remains the data source. The presentation is built around my question.
Many individual records become a view that helps me understand and manage our business better.
What sells at the bar — and where can we help?
This also gives me a different perspective on our bar data. I can examine revenue and drinks sold by bartender, shift and team composition. Which products sell particularly well? Which combinations of team members stand out for good results? Where is a closer look worthwhile?
Our premium gin is a practical example. Is it being actively recommended and sold? Are there differences we can learn from? Does someone need more product knowledge to recommend it confidently to guests?
These analyses provide starting points for conversations and support within the team. The event, guest numbers and the bar someone worked at also matter when interpreting the results.
For me, the value comes when the numbers lead to a concrete decision: explain a product better, share experience across the team or monitor a trend more closely.
Suddenly, Salesforce is above the bar
My personal favourite is our drinks menu. I had a public artifact created for the screens above our bars: an animated menu using drinks and prices from Salesforce. It runs as a website in full-screen mode on our displays.
When prices change or new items are added, I can have the updated information from Salesforce loaded into the artifact. I no longer need to design and export a new video.
That is a lovely moment: data we already maintain to run the business becomes directly useful to our guests. This makes the headless approach very tangible. I work with Salesforce through Claude and have an interface created for a specific purpose. Salesforce remains the data source, while its use extends all the way to our bar.
What I have learned along the way
1. Coworker is my starting point and my orchestrator
For me, Coworker is a quick and accessible way to start using AI as a colleague in my Salesforce org. I describe the task, and Coworker looks for a way to carry it out. I see Coworker as a complement to specialised Agentforce agents. My shift-creation agent is a good example: its capability remains valuable. Coworker can use it in the right context and gather the necessary information beforehand. Within my Salesforce instance, Coworker is becoming a kind of “superagent”: a central point of contact that handles tasks using its own capabilities or brings in a specialised agent. This adds value to the work I have already invested in my agents.
2. An artifact needs an owner, too
Claudeforce makes it easy to develop your own analyses and interfaces. But if everyone now builds their own revenue dashboard, we will soon have five nice-looking views and five different results. We will have used AI and created our next problem in the process. One person includes events that have not yet been finalised; another excludes them. One compares calendar dates; another compares weekends. There is also token usage and duplicated effort in development, checking and maintenance.
For me, a good artifact therefore needs clear ownership. Who is responsible for the calculation logic? Who checks changes? Which version do we use together?
Artifacts can be organised and shared. I would work out how that should function within the team before the first artifact goes into operational use, wherever possible. That creates something people can use and develop together.
3. Good results start with a clear description
My experience is that the more precisely I describe the task, the better the result.
That takes more than “Build me a dashboard”. I need to explain which data sources matter, what individual fields mean and how the information should be interpreted.
For our year-on-year comparison, for example, it is essential to compare a Friday with a Friday. How to handle events whose accounts have not yet been finalised is equally important. I need to state these business rules explicitly.
It is also worth thinking about the presentation. Do I want to identify trends over time, compare individual events or understand where our money goes? Each calls for a different view. So I try to describe as concretely as possible which question I want to answer and what I intend to do with the result. That preparation pays off in the quality of the artifact.
4. The first version is the beginning
For all my enthusiasm, a good result still takes time, even with AI. Sometimes it is only the first version that reveals a missing requirement. A visualisation looks good but does not yet answer the actual question. A calculation needs to be refined. Or conversations with users reveal that they need a different view. For me, this exchange is part of development: try things out, check the results, gather feedback and refine them together. With each iteration, I get better at describing what I need. At the same time, the tool becomes a better fit for our day-to-day work.
From an assignment to the bar
The foundation for all these examples also lies in the work we did beforehand: team members are assigned, Slack IDs are stored, events and shifts are linked, and supporting documents and payments are recorded in a structured way. Coworker and Claudeforce build on that foundation.
For me, this now means six to ten fewer hours of routine work each month and a better understanding of our business. The information was there before. Now I can actually put it to use in my day-to-day work.
Conclusion
Our nights may be longer and our outfits occasionally shorter, but when it comes to staffing, finances and informed decisions, we face the same challenges as other businesses.
For me, this shows where AI’s strategic value lies: bringing together existing data, processes and specialised agents so the business can act on them. With clear accountability and solutions that people use together.
Saving time is the beginning. The bigger gain is a business that turns what it knows into action faster.
