SELLSATION AI Sales Assistant – Administration

The next best action configuration has three tabs.

Picture 1 – The three tabs of the next best action configuration

No.	Tab	What is set there
1	Next best action	The AI Sales Assistant proposes the actions that lead to a close. Here you decide which ones exist and when they appear – drafted by the AI or written yourself.
2	Close probability and scoring	The two figures of a project – how likely the close is, and how important the project is. These values too the AI works out for you, or you set them yourself.
3	Preview	Try the settings on real projects without anything being triggered

Two figures stand side by side in the assistant and are easily confused.

• The close probability says how likely the project is to be won.

• The deal score says what priority the project has.

Picture 2 – Close probability and deal score side by side – the same in the AI Sales Assistant in the detail view as in the overview

The deal score starts at the close probability and carries on from there: points added for what is going well, points taken off for what is missing. The result becomes the level of the project – Critical, High, Medium, Low (section 3.3).

An example: two projects at 60 % close probability, an offer sent on both. Project B is also missing the decision maker, and there has been no contact for 40 days.

	Project A	Project B
Starting figure from the close probability	60	60
Offer sent	+6	+6
No decision maker identified	–	−8
Long inactivity	–	−10
Deal score	66	48

Same close probability, different priority.

	Close probability	Deal score
Scale	0 to 100 percent	0 to 100 points
Appears in	project and assistant	assistant only

To be distinguished from that is the priority of a single recommendation: it says which step is next. A project with a low deal score can carry a critical recommendation.

Picture 3 – A single recommendation with a priority of its own, here High in a project with a low deal score

1. General

Picture 4 – The Next best actions configuration page in the administration under Apps & integrations

The AI Sales Assistant proposes what to do next and rates every project.

Picture 5 – The upper part of the Next best action tab with its buttons

On the Next best action tab all recommendations stand one below the other, in the order in which they are checked. Below each name is who created it and when somebody last changed it – for recommendations from the AI as well.

Above them are these buttons:

No.	Button	What it does
1	Take over the standard settings	Ticking the box writes the recommendations that ship with the product down as configuration. After that you change them like any of your own – triggers, conditions, actions.
2	Let the AI work some out	The assistant proposes new recommendations out of your closed projects (chapter 5)
3	New next best action	Write a recommendation of your own
4	Let AI set the order	Reorders the list – by scoring, by close probability or by the projects examined

The tick box deletes nothing. Take it away and the recommendations taken over are merely switched off and hidden – ticking it again brings them back with your changes on them. If you have deleted them, only Restore deleted configurations brings them back.

Everything is set in the administration under Apps & integrations on the Next best actions configuration page: the recommendations themselves (chapter 2), the close probability (3.1) and the deal score (3.2).

Everything can be set. Changes act on both views: on the assistant in the project detail view and on the overview.

Only administrators see these pages. An employee sees the result but cannot change it.

This article assumes that you know the two user articles – SELLSATION AI Sales Assistant in the project detail view and SELLSATION AI Sales Assistant in the overview. They describe how the settings show up for the employee.

2. The Next best action tab

Picture 6 – One entry of the list with its controls

No.	Element	What it does
1	An entry of the list	A recommendation, folded up
2	Switch Active	Switches this recommendation on or off without deleting it
3	Arrows ↑ ↓	Move the entry up or down the list
4	Menu ⋯	Duplicate and Delete

2.1. What a recommendation holds

The arrow to the left of the name unfolds a recommendation. From top to bottom:

The texts. What the employee sees on the card is highlighted in blue.

Picture 7 – The texts: Name, Reason, Suggested actions, and below them what the figures advise

When the recommendation appears.

Picture 8 – Triggers, conditions and exclusions

No.	Block	What it is for
1	Trigger	What raises the recommendation at all. Here: Project is open
2	Additional conditions	Narrow down where it appears. Here: Time pressure present is No
3	Exclude when	Keep it away
4	Applies to	Which projects it applies to at all, and how long it stays hidden once it has been done

Whether a recommendation turns up is decided on three levels – linked differently:

Level	Link	Effect
Triggers	OR	One is enough. Each trigger on its own can raise the recommendation.
Conditions	AND	All of them have to hold, otherwise nothing appears.
Exclusions	AND NOT	If one holds, the recommendation stays away.

So entering several triggers makes the recommendation appear more often, not less. To narrow it down, use the conditions.

What the recommendation offers. The button the employee presses on the card.

Picture 9 – The action behind the recommendation

No.	Block	What it is for
1	The action to run	The kind of action and a guide figure in minutes. Here: Create an appointment
2	Action details	The fields that belong to the chosen kind of action
3	Automation	Whether a task is created as well, with a due date and someone responsible
4	Who it is aimed at	Whom the action is directed at
5	Audience	Who gets to see the recommendation

What it brings about. The two figures the recommendation promises.

Picture 10 – What the recommendation is worth

No.	Block	What it is for
1	Moves	Which close probability factor this step clears or brings about
2	Worth	By how much close probability and deal score rise once the step is done

A recommendation holds, among others, these entries:

Entry	Meaning
Name	The headline of the card, for example „Identify the decision maker"
Reason	Why this recommendation stands on this particular project
Suggested actions	Two or three sentences on what exactly to do
Applies to	Open, won, lost or all projects
Base score	The starting figure of the score (section 2.5)
Moves	Which deal score factor this step clears or brings about
Worth	By how much chance and deal score rise
Estimated time	Guide figure in minutes; left empty, the kind of action decides

Every one of these entries can be set. At Worth there is also Recalculate with AI: the assistant counts out your closed projects and proposes both values. What you enter yourself stays as it is.

The three text fields are read by the employee on the card unchanged. Write them so that they are understandable without your settings – the rule behind them is never shown.

Moves is the entry most easily overlooked. It connects the recommendation to the deal score: only through it does the assistant know that this step clears a particular risk – and can work out what it is worth.

2.2. The triggers

SELLSATION ships ready-made triggers, grouped by topic. The list is not exhaustive and grows with the program – at the moment, for example:

Group	Examples
Communication	No communication for X days, no e-mail, no answer, no call
Project	X days in the same phase, decision date overdue, chance or deal score below/above X, amount above a figure
Offer and competition	Offer unanswered for X days, offer rejected, a competitor involved
Tasks and appointments	Task overdue for X days, no appointment in the future, no follow-up
Buying center	No decision maker recorded, executive level not involved
Read out of the history	Budget named, a presentation took place, references sent

Can be set: which triggers you use and with what value – 30 days instead of 14, say.

If the list is not enough, build one of your own out of three parts: source (what is looked at), question (since when, is above, equals …) and value.

2.3. Where the answer comes from

Beside every trigger it says what it rests on – that tells you how reliable it is:

Source	Meaning
Data	Straight out of the record. Unambiguous.
Read out of the history	Taken from the text of the activities. A good guess, not a proof.
Mixed	The record first, and where it is silent, the history.

Cannot be set – the source belongs to the trigger. You can decide, though, whether you want to rely on a trigger that is only read out of the history.

2.4. What a recommendation offers

Every action leads to a concrete next step. On offer at the moment, among others:

Kind of action	What happens
Write an e-mail	Opens the e-mail window with a prepared subject
Write an e-mail with the assistant	The AI drafts it – you give purpose, tone and instruction
E-mail from a campaign template	Uses a campaign template that already exists
Create a call task	Creates a task of the call kind
Create an appointment	Opens the appointment window
Create a task	Creates a task, with title and description from what you set
Activity	Something done rather than sent – work out a price, prepare an argument
Assign roles to contacts and create new ones	Leads to the people on the project: hand out roles, add the contacts who are missing
Write a note	Records a note, nothing else
Fill in a field	A project field is asked for on the card itself and saved there
Move phase	Moves the project into another phase

Depending on the kind of action you also set, for example:

Setting	On which kinds
Recipient – main contact, decision maker, whole buying center	e-mail, appointment
Responsible – customer consultant, project consultant, a fixed user	task, call, activity
Due date and reminder	task, call, appointment
Visible to	all
Purpose, tone and instruction for the AI	e-mail with the assistant
Propose attachments	the e-mail kinds
Which field is asked for	fill in a field
Estimated time	all

Two traps:

• At the reminder, empty means no reminder – that is something other than zero days before.

• The estimated time you usually leave empty; then the kind of action supplies its guide figure. Enter something only where the kind of action does not describe the work – writing an offer is an hour, the activity for it five minutes.

Where a recommendation holds, the employee gets buttons with which to carry it out straight away.

Can be set: which actions are offered and in what order. A recommendation may offer several – the first one is highlighted.

2.5. The score of a recommendation

Every recommendation has a base score. On top of it come the points of its influence factors, as far as those hold. That gives the order in which the employee sees them.

The recommendations that ship carry different base scores: Draw up the offer and place it stands far up, Send references further down.

Can be set: the base score of every recommendation and the points of its factors. The direction you choose beside it as positive or negative; a typed minus sign does not exist.

3. The Close probability and scoring tab

Picture 11 – The Close probability and scoring tab

No.	What it is
1	The buttons of the tab – standard values, AI calculation, bin
2	The deal score block: starting figure, switch Start from the opportunity, AI proposal for every value
3	An unfolded factor with its three groups of fields
4	A folded factor – Scoring and Close probability stand side by side

3.1. The close probability

The close probability is the percentage that says how likely the win is. It is set in two places:

Where	What you set there
Close probability and scoring, column Close probability	By how many percentage points a factor raises or lowers the chance
On the single recommendation, field Worth	By how many percentage points the chance rises once the employee has done this step

Two figures per factor

The same list of factors carries both values side by side.

Column	Acts on	Example
Scoring	the deal score	+6 points
Close probability	the percentage of the chance	+4 %

The direction holds for both: a factor of the negative kind takes off points and percent.

The fields of a factor are split into three groups:

Group	Fields
What the factor is	Factor – the trait itself; Outcome – how it appears in the Moves picker
When it holds	Condition / threshold, Comparison, Threshold
What it is worth	Scoring, Close probability, Kind (positive or negative)

Beside most fields is AI suggestion – with it the assistant fills the single field instead of working out every value at once.

Empty is not zero. If the Close probability column stays empty, this factor leaves the chance as it is – nobody has rated it. An entered 0, by contrast, means: checked and without effect.

Can be set: both figures per factor. With AI suggestion for every value the assistant works them out from your won and lost projects and then keeps them current nightly – below it stands how many values that concerns. What you type in by hand is left alone.

The deal score and the close probability in the project detail view can differ slightly. The detail view also takes into account how often there was a conversation and how often the project was moved.

3.2. The deal score

The deal score is the priority of the project. It rates the project as a whole on a scale from 0 to 100 and gets the same levels out of it as a single recommendation: Critical, High, Medium, Low.

Not to be confused with the priority of a single recommendation. That says which step is next. The deal score says which project deserves your time at all. All recommendations on a project share the same deal score.

It starts at the close probability of the project – that is what the switch Start from the opportunity does. If a project carries no chance of its own, the base score applies instead. Points are then added and taken off, according to what holds on the project.

Leave the switch on. Without it every project starts at the same base score, and the deal score then differs noticeably from the close probability in the detail view.

Can be set: points and direction of every factor, and you can write factors of your own – out of field, comparison, score and direction. Yes/no entries and counts are possible, time spans are not. Single factors can also be switched off or deleted.

What speaks for it:

Factor	Points
Offer sent	+6
Meeting in person held	+5
Presentation held	+4
Executive level involved	+4
Several decision makers involved (from 2)	+4
References sent	+3
Budget named	+3
Supporter present	+3
Need confirmed	+3
Time pressure present	+3
Decision process known	+3
Offer drawn up	+3
Buying center staffed (from 3 contacts)	+3

What speaks against it:

Factor	Points
Competing offer open	−12
Long inactivity (from 30 days)	−10
No decision maker identified	−8
Overdue tasks open	−6
Price resistance	−6
Requirement not covered	−6
Procurement or legal blocking	−6
Competitor has quoted	−6
Decision date missing	−5

Two things on that:

• Plus and minus count the same. A factor acts with exactly the points you enter – whether it adds or takes off.

• The figures are starting values. What a factor is really worth on your data, the assistant works out from your closed projects (chapter 5).

3.3. The levels

The finished score becomes a level – what the employee sees as Critical, High, Medium or Low.

Picture 12 – The Priority levels block on the Close probability and scoring tab

Level	from (recommendation)	from (project)	Colour
Critical	80	75	Red
High	70	60	Orange
Medium	55	40	Blue
Low	0	0	Grey

The two columns are needed because two different figures fall into the same level: the score of a single next best action and the deal score of the whole project.

Can be set: name, colour and both thresholds of every level – on the Close probability and scoring tab right at the bottom, below the deal score block. You can write levels of your own and delete existing ones.

• The lowest level catches everything else. Its threshold belongs at 0. The Critical and High priority tiles in the overview always count the top two levels – rename them and the tiles change with them.

4. The Preview tab

Picture 13 – The Preview tab after a run on real projects

No.	What it is
1	The note that nothing is carried out and nobody is written to
2	The controls: how many projects, how far back – and the button that starts it
3	The result in figures: projects examined and recommendations raised
4	A result card with level, score and the why

If that gets you no further, the Preview tab helps. It tries the settings out on your real projects – as they stand on screen, saved or not.

Can be set: how many projects are examined and how far back events count. Afterwards it says how many recommendations were raised, with a why for each one. If nothing is raised, the preview advises widening the timespan of a trigger or taking a condition out.

Nothing is carried out and nobody is written to. The preview only calculates.

5. What the assistant proposes by itself

The assistant evaluates your closed projects and proposes:

Proposal	What is behind it
New recommendations	States in which your projects are won less often
Point values	What a factor is really worth on your data, instead of the starting value that ships
Traits of won deals	Fields that won projects had filled in and lost ones mostly did not

Can be set: whether you take a proposal up. The assistant changes nothing by itself.

5.1. Counted, not guessed

This is arithmetic, not a language model: the assistant counts how often a condition held and how many of those projects were won.

Beside every finding it says what it was counted from: how many projects were concerned, how many of those were won, and how that stands against the usual win rate.

5.2. When the assistant stays silent

Hurdle	Value
Closed projects per outcome	at least 150 won and 150 lost
Projects per condition	at least 25
Drop in the win rate	at least 5 percentage points

If your tenant has too little history, the assistant says so and proposes nothing.

5.3. What it leaves out

• Only the negative. It looks only for states in which deals are won less often.

• Chance, deal score and status do not count – the outcome writes those itself.

• Order number, invoice number, win date and names stay out – they only exist after the decision.

5.4. Traits of won deals

The recommendation Capture the traits of won deals does not ship with the product, it is produced from your projects. Below every field it says what won projects had entered there.

5.5. Whom a figure belongs to

Values set by the assistant are marked, and it keeps them current. As soon as you type into the field, the figure is yours and is no longer overwritten.

6. Back to the starting state

Everything you change can be taken back. The buttons for it do not do the same thing:

Button	Where	What it brings back
Restore the standard settings	on the single recommendation	This one recommendation to the state it ships in – trigger, conditions, actions, everything
Take over the standard settings	Close probability and scoring tab	Points and direction of the factors to the values that ship
Put the hand-typed figures back	Close probability and scoring tab	The values that stood there before the AI calculation
Restore deleted configurations	in the list	Deleted recommendations – the ones that ship as well as your own

Careful, two different things carry the same label: Take over the standard settings on the Next best action tab brings in the recommendations that ship (chapter 1); the same label on the Close probability and scoring tab resets the points.

Three notes on that:

• Changes are visible. Above a changed standard recommendation stands a note with a button to reset it. On the factors it also says who last changed them and when.

• One exception: Take over the standard settings resets points and direction, but not the close probability – the product ships no figure for that.

• Deleted does not mean gone. What an employee deletes in the overview also stays available under Deleted next best actions.

Can be set per recommendation: how long it disappears once it has been done. Empty means 90 days, 0 means permanently – until somebody brings it back under Completed next best actions.

7. Requirements and notes

7.1. Requirements

The AI Sales Assistant has to be switched on for your tenant, and you need administrator rights. Without one of the two you do not see the pages.

7.2. When a change arrives

Not at once – the result is kept for 15 minutes. To see your change right away, press the refresh arrow next to Last updated in the overview. After larger changes the first calculation takes a few minutes.

7.3. Texts the employee reads

What you write into the text fields appears on the card unchanged – in your language, none of it is translated. Which areas the employee sees at all you recognise by the blue highlighting.

7.4. What cannot be set

Two things, so that you do not go looking for them:

• The triggers Offer not opened and Buying center unchanged do not exist – the data needed for them is recorded nowhere.

• A deal score factor of your own cannot measure in days or frequencies.

7.5. If a recommendation does not appear

Work through this order:

1. Is it active?

2. **Does Applies to fit?** A recommendation for won projects never appears on open ones.

3. Is a condition too narrow? Conditions are linked with AND – a single one that does not hold is enough.

4. Does an exclusion apply?

5. Has it been recalculated yet? See 7.2.